Results

Between 1979 and 1999, overweight and obesity prevalences increased from 16% to 21% and from 2% to 4% (table 1). The annual increase of overweight and obesity was 0.4% and 0.2%, respectively. The inter-period correlation was high, tending to support the stability of municipal rates between periods (inter-period correlation for overweight is 0.72, 0.7, 0.7  for one, two and three period lags, respectively; for obesity prevalence, these coefficients were 0.56, 0.57, 0.55, respectively).

The Robin Hood Index indicated that, in 1990, about 9% of overweight and 19% of obesity would have to be redistributed from high prevalence areas to low prevalence ones to achieve an equal distribution among municipalities. This inequality pattern remained unchanged over the entire period. In 1990, the unequal socioeconomic distribution of obesity gave poorer municipalities 3% more overweight and 8% more obesity, compared with better off areas. Thus, socioeconomic inequality of obesity and overweight seems to be more variable but neither increases or decreases.

Discussion

Overweight and obesity prevalences among young Belgian men present a slight geographical inequality showing no significant change throughout the 1980s. As obesity increased during the same period, this implies that the whole country shared this higher prevalence burden.

Highly educated men were less likely to undergo military service. During the 1980s, measures were taken by the Defence Secretary to enforce a more equal enrolment rate among different educational groups. To the extent that body mass index decreases with the socioeconomic status, lower proportion of enrolment of highly educated men might lead to a slight overestimate of overall prevalence. The enforcement of the rules may have changed the socioeconomic pattern of examined men during the period considered. However, the relation between the socioeconomic status and body mass index seems to be more relevant for women than for men.5

Lower health status might have decreased enrolment and increased body mass index, hence producing a downward bias of obesity prevalence. Nevertheless, as our database refers to examined population, whether enrolled or not, our results are less vulnerable to such a bias: obese young men may not have been enrolled, but after the clinical examination.

Examination and enrolment rate was higher in the north of the country and lower in the south. As the French speaking region of the country is known to have a lower nutritional status,6 this may lead to underestimation of obesity prevalence and of geographical inequality.

The Robin Hood Index is poorly suited for transfer of obesity/overweight between two municipalities located on the same side of the mean. However, the Lorenz curves are mostly parallel (not shown here), which means that the overall distribution is stable in time: hence, the use of such an index is appropriate.

While a global and coherent strategy to tackle obesity is emerging,7 this work highlights the need for monitoring and strategies taking into account the geographical inequality of such a risk factor. It is, in fact, very likely that other cardiovascular risk factors may show similar geographical inequality. WHO Cardiovascular Monitoring projects have been developed in Belgium since 1985 in three areas. They can be helpful to track such inequalities.