Influence of dietary factors on weight change assessed by multivariate graphical models

To analyse the influence of dietary factors on weight changes in a multifactorial context by use of graphical models.

Two age-stratified cohorts of Danish men and women were examined twice with five years interval.

The Glostrup Population Studies in the Copenhagen County, Denmark.

A random population sample of 2009 men and women, aged 30, 40, 50, and 60 years.

At baseline all underwent a general health examination including measurement of height and weight, and they completed questionnaires regarding familial predisposition, cohabitation, school education, vocational training, social class, smoking habits, physical activity during work and leisure time, and parity. Dietary intake was evaluated by 7-day food records, allowing estimation of total energy intake and intake of fat and alcohol. At the follow-up examination height and weight were measured again and at each examination BMI was calculated. The statistical analysis was performed by graphical models based on multidimensional contingency tables defined by independence graphs, these may include several categorial or ordinal scale variables and include the assumed temporal relationship between the variables.

BMI at entry depended on familial predisposition to obesity, and for women on cohabitation as well. At follow-up 5 years later, BMI in the male group was only related to former BMI, whereas in the female group it was influenced by parity as well. There were several significant associations between the dietary factors, other life style habits and social factors. However, no significant associations were found between total energy intake, percentage of energy from fat or from alcohol and weight changes. For neither gender was there any significant influence of other putative determining factors.

No influence of dietary factors on weight changes were found when assessed in the context of many other putative determinants of weight change.