Some statistical aspects of food intake assessment.

OBJECTIVE: To present the results of the statistical working group of the EFCOSUM project on estimating the minimum sample size for a pan-European dietary survey. BACKGROUND AND METHODS: Numerous statistical issues are involved when planning a nutritional survey aimed at evaluating various indicators, especially if it will be carried out in different countries. The plenary workshop of the EFCOSUM project has chosen four relevant statistical topics: the sample size estimation for dietary surveys, the number of repeated measurements needed to estimate usual intake for each individual; the statistical presentation of data; and the statistical procedures for estimating the usual intake distribution from a limited number of days of observation. This article deals with the first three topics mentioned. The participants of the EFCOSUM project answered a small questionnaire in order to get agreement on the method of estimating a minimum sample size in the context of a monitoring of dietary indicators. Data on the variability of dietary indicators of interest was also collected, in order to calculate a minimum sample size. RESULTS AND CONCLUSION: The main result was that a minimum sample size of 2000 adults in each European country will be needed in order to identify trends in the mean intake of the most relevant foods and nutrients in Europe. This sample size should be higher if trends have to be indentified for socio-demographic subgroups.

Analytical approaches to food-based dietary guidelines in the European setting.

OBJECTIVE: To show the effects of statistical approaches of data analysis to be used in the development of Food-Based Dietary Guidelines (FBDG). SETTING: Databases from dietary surveys in 6 European countries. RESULTS: Quantile analysis based on iron intake among adult women resulted in differences among European countries regarding (macro) nutrient intake and consumption of food groups. However, in all countries women in the highest quartile had a higher intake of energy and dietary fibre and a higher intake of most food groups. In developing FBDG adjustment of energy intake is recommended. Discriminant analyses showed that among Dutch women potatoes, red meat, sausages, offal, savoury snacks, eggs and total vegetables were found to be the most predictive for differences in iron intake. Relatively high correlations were observed for iron and dietary fibre and iron and (some) B-vitamins. Examples from cluster and factor analysis showed that this type of analysis considers the complexity of the dietary pattern and could also be a helpful instrument in the development of FBDG. CONCLUSIONS: The use of a nutrient distribution can be used as a minimum approach in developing FBDG. More advanced methods can also be used in addition to set priorities for FBDG and to analyse complete dietary patterns.

Under-reporting in dietary surveys--implications for development of food-based dietary guidelines.

OBJECTIVE: To illustrate issues related to under-reporting in dietary surveys relevant for development of food-based dietary guidelines. SETTING: Food and nutrient intakes in two national dietary surveys are analysed in relation to under-reporting using the concept of cut-off values for the ratio between energy intake and basal metabolic rate as well as biochemical markers of nutritional status. RESULTS: Subjects with an implausibly low EI/BMR ratio had dietary patterns more close to recommended levels, e.g. lower fat intake and higher fibre intake, than those with acceptable energy intakes. Intakes of micronutrients were lower. This was due to both general and selective under-reporting of foods. Under-reporting was due to less frequent consumption, larger proportion of non-consumers and smaller portion sizes. Low EI/BMR ratios did not influence the characterisation of protein and micronutrient intake of low and high consumers of fruit and vegetable intake. CONCLUSIONS: Under-reporting can influence the usefulness of dietary data as a tool for developing food-based dietary guidelines, and the extent of under-reporting should be assessed using appropriate methods.

Nutrient intake of physically fit and unfit men and women.

OBJECTIVE: Participation in physical activity and following a diet low in fat, saturated fat, and cholesterol and high in fiber are important in the prevention and treatment of cardiovascular disease and other health problems. Despite the importance of both behaviors, little is known about the interactive role of diet and physical activity. The association between physical activity and diet has been studied, but data on the association between cardiorespiratory fitness (CRF), which is an objective measure of habitual physical activity, and diet are lacking in adults. This report examines nutrient intakes of men and women across low, moderate, and high fitness categories and compares the intakes to national dietary recommendations. It is the first step in examining the relationship between diet, CRF, and morbidity and mortality endpoints in the Aerobics Center Longitudinal Study (ACLS). METHODS: Between 1987 and 1995, 7959 men and 2453 women participating in the ACLS provided 3-d diet records and completed a preventive medical examination. CRF was measured using a maximal exercise test, anthropometric and other clinical variables were measured following a standardized protocol, and lifestyle factors were assessed with a medical history questionnaire. RESULTS: After adjusting for potential confounders, there was a significantly lower percent of energy from fat and saturated fat across low, moderate, and high CRF categories. With additional adjustment for total energy intake, there was a significantly higher dietary fiber intake and a significantly lower cholesterol intake across CRF categories. The percentage of men and women meeting national dietary recommendations was higher at higher CRF levels. CONCLUSIONS: Men and women with higher fitness levels consumed diets that more closely approached national dietary recommendations than their lower fit peers.

Comparison of a dietary record using reported portion size versus standard portion size for assessing nutrient intake.

OBJECTIVE: Because the percentage of missing portion sizes was large in the Aerobics Center Longitudinal Study (ACLS), careful consideration of the accuracy of standard portion sizes was necessary. The purpose of the present study was to investigate the consequences of using standard portion sizes instead of reported portion sizes on subjects' nutrient intake. METHODS: In 2307 men and 411 women, nutrient intake calculated from a 3-day dietary record using reported portion sizes was compared with nutrient intake calculated from the same record in which standard portion sizes were substituted for reported portion sizes. RESULTS: The standard portion sizes provided significantly lower estimates (>/= 20%) of energy and nutrient intakes than the reported portion sizes. Spearman correlation coefficients obtained by the two methods were high, ranging from 0.67 to 0.93. Furthermore, the agreement between both methods was fairly good. Thus, in the ACLS the use of standard portion sizes rather than reported portion sizes did not appear to be suitable to assess the absolute intake at the group level, but appeared to lead to a good ranking of individuals according to nutrient intake. These results were confirmed by the Continuing Survey of Food Intake by Individuals (CSFII), in which the assessment of the portion size was optimal. When the standard portion sizes were adjusted using the correction factor, the ability of the standard portion sizes to assess the absolute nutrient intake at the group level was considerably improved. CONCLUSIONS: This study suggests that the adjusted standard portion sizes may be able to replace missing portion sizes in the ACLS database.

Longitudinal development and tracking of calcium and dairy intake from teenager to adult.

OBJECTIVES: Because of growing interest in the potential role of calcium in preventive pediatric strategies against osteoporosis, the longitudinal development and tracking of calcium had a dairy intake from adolescence into adulthood was addressed. DESIGN: In the Amsterdam Growth and Health Study, a group of 84 males and 98 females were followed over a 15 y period from age 13-27 y. The calcium and dairy intake was assessed six times by a cross-check dietary history method. To assess tracking, two traditional approaches, interperiod correlation coefficients and percent of subjects who remained in the same quartile of calcium intake over time, and a new approach based on generalized estimating equations were used. RESULTS: The mean calcium intake was relatively high and increased over time (30%). The tracking of calcium and dairy intake from adolescence into adulthood was moderate in both sexes (for example the correlation for calcium was 0.43 in males and 0.38 in females). CONCLUSIONS: The predictability of calcium intake over time does not seem to be sufficiently strong to identify teenagers who are likely to maintain an inadequate calcium intake in adulthood. Therefore, the identification and treatment of subjects with a low calcium intake cannot be limited to the teenage period but should be extended into adulthood.

The Amsterdam Growth and Health Longitudinal Study. The past (1976-1996) and future (1997-?).

This article presents an overview of the Amsterdam Growth and Health Longitudinal Study (AGAHLS). This study was started in the 1970s, following a whole sample of 500 healthy 13-year-old boys and girls from two secondary schools. During the school period (12-17 years) annual measurements were performed with respect to anthropometrics, physiological and psychological parameters, lifestyle characteristics (activity, diet, smoking) and health parameters. A multiple longitudinal design was applied with overlapping birth cohorts and a cross-sectional measured control group to monitor for confounding factors such as time of measurement, cohort, dropout and testing effects. Emphasis is also placed on measures that enhance adherence of the subjects. The follow-up was extended with repeated measurements at age 21, 26 and 28. This enabled us to analyze the data with respect to tracking characteristics of biological and lifestyle variables over a period of 15 years between adolescence and adulthood and also to investigate quasi-causal relationships between the effects of a healthy lifestyle and indicators for cardiovascular diseases (CVD). Also new methods such as measurement of atherosclerosis and osteoporosis will be added and related to longitudinal measurements of the same subjects in the past. The main results that were obtained in the past are summarized and research questions for the near future are explained.

Relative validity of 16-year recall of calcium intake by a dairy questionnaire in young Dutch adults.

Assessing past calcium intake in studies of osteoporosis is of major importance because calcium intake during young adulthood is thought to influence bone mass in later life. However, uncertainty exists about the accuracy of retrospectively reported calcium intake. We therefore evaluated whether a dairy questionnaire estimating calcium intake only from dairy products is a valid method to assess the retrospectively reported calcium intake of 29-year old Dutch males and females 8 and 16 y previously. This was examined in 166 subjects by comparing their original calcium intake reported in 1977 and 1985 by a cross-check dietary history method, with their 1977 and 1985 calcium intake values recalled in 1993 by the dairy questionnaire. The overestimation of the retrospectively reported calcium intake compared with the original intake (11-33%) suggests that the dairy questionnaire is a relatively poor tool to assess the absolute calcium intake of young adults 8 and 16 y ago. Based on the agreement statistics, the dairy questionnaire is a reasonably valid instrument to rank subjects according to their retrospective calcium intake. However, the reporting of the retrospective calcium intake appears to have been influenced by the current calcium intakes reported at the age of 29 y. Nevertheless, the retrospective calcium intake appears to be a better indicator of the original intake than the current intake because the subjects markedly increased their calcium intake over the 16-y period.

Biological maturation and the distribution of subcutaneous fat from adolescence into adulthood: the Amsterdam Growth and Health Study.

OBJECTIVE: To analyze differences in the development of a trunk-oriented fat distribution pattern between 13 and 27 years of age in individuals who either matured rapidly or slowly in adolescence. DESIGN: Longitudinal, observational, four annual measurements between 1977 and 1980 and additional measurements in 1985 and 1991. SUBJECTS: 79 boys, 98 girls, health Caucasian schoolchildren, classified as rapidly, normally or slowly maturing. MEASUREMENTS: biceps, triceps, subscapular, suprailiac skinfolds resulting in two skinfold ratios, body mass index, skeletal age, peak height velocity (for boys only), age at menarche. RESULTS: No statistically significant differences were found between rapidly and slowly maturing boys and girls, based on the skeletal age or the peak height velocity. Girls with a relatively early menarche showed significant higher mean skinfold ratios between 13 and 27 years of age than girls with a relatively late menarche. CONCLUSION: Skeletal maturation of boys and girls and peak height velocity (only measured in boys) are not associated with a trunk-oriented fat distribution pattern between 13 and 27 years of age. Only a relatively early menarche in girls seems to be associated with a trunk-oriented fat distribution pattern from adolescence into adulthood.

A meta-analysis of the effect of calcium intake on bone mass in young and middle aged females and males.

Evidence of a direct relationship between dietary calcium and bone mass in young adults is inconsistent. Therefore, a meta-analysis was performed to evaluate the existing literature concerning this relation in premenopausal women and adult men between the ages of 18 and 50. A computer-aided search of published papers from 1966 through 1994 was conducted. The quality of each study was assessed by examining the study design. The results of the well-designed studies wer combined to obtain quantitative conclusions. A total of 33 eligible studies were identified in the literature, representing 27 cross-sectional studies, two longitudinal and four intervention studies. The results of the cross-sectional studies in permenopausal females showed a significant correlation coefficient (r = 0.13, 95%-confidence interval = 0.09-0.16) as well as partial correlation coefficient (partial r = 0.08, 95%-confidence interval = 0.05-0.12), both weighted by sample size. The intervention studies found that calcium supplementation of approximately 1000 mg/d in premenopausal women can prevent the loss of 1% of bone/y at all bone site except in the ulna. In conclusion, the studies published to date seem to offer overall evidence that calcium intake is positively associated with bone mass in premenopausal females. This association is fairly consistent across the different study designs and is strengthened by the fact that the results are based only on studies with a high methodological quality. In males, too few studies (only three) were published to draw firm conclusions.

Comparison of a quantitative dairy questionnaire with a dietary history in young adults.

BACKGROUND: There is growing interest in the potential role of calcium in the prevention of osteoporosis. Therefore a quantitative dairy questionnaire (DQ) was developed to estimate the calcium intake from dairy products. The aim of this study was to evaluate the relative validity and reproducibility of the DQ in 27-29 year old males and females. METHODS: The DQ was validated against a cross-check dietary history method (DH) in 160 subjects, and the reproducibility was assessed in 29 subjects by repeating the administration of the DQ with a one-year interval. RESULTS: In the validity study, the nutrient calcium and the food groups cheese and milk and milk products measured by the DQ tend to show higher intakes as compared to the results of the DH. However, only for the calcium intake was the difference between the DQ and DH (7%) significant. The Bland-Altman plot showed that this bias was constant over the range of calcium intakes. The Pearson correlation coefficients between the DQ and the DH varied from 0.58 to 0.65. On average, about 52% of the subjects were classified in the same quartile in the DQ and the DH for the intake of calcium as well as dairy products. The weighted kappas were between 0.60 and 0.67. In the reproducibility study, the Pearson correlation coefficient for the calcium intake assessed by the DQ one year apart was 0.78. CONCLUSIONS: These data indicate that the DQ is reasonably accurate in young adults as a means of assessing the calcium intake from dairy products.

Weight-bearing activity during youth is a more important factor for peak bone mass than calcium intake.

In a 15 year longitudinal study (Amsterdam Growth and Health Study) is evaluated the effect of daily calcium intake (CAI) during adolescence and young adulthood on the development of peak bone mass at age 27 when the influence of weight-bearing activity (WBA) and body weight was accounted for. A group of 84 males and 98 females were measured longitudinally from age 13 until age 28. Measurements were taken six times of anthropometric characteristics. Lifestyle was also evaluated six times by cross-check interviews of CAI and WBA. Bone mineral density (BMD) of the lumbar spine was determined at age 27 by dual x-ray absorption. Three periods were considered, that is, the adolescent period, the period 13-21 years, and the total period (13-27 years). In multiple linear regression analyses, only WBA and body weight were significant positive contributors to the final model of lumbar BMD at age 27. In all three periods WBA was the best predictor in males and body weight in females. ANOVA was performed on BMD and the highest and lowest quartiles of calcium intake with the significant predictor variables of the linear regression model as covariates. Again calcium intake appeared not a significant predictor of BMD in the three periods in both sexes. Regular weight-bearing exercise and at least a normal age-related body weight in adolescence and young adulthood are of key importance in reaching the highest lumbar peak bone mass at the age of 27 years.