Basic descriptive statistics for males and females at each date are shown in Table 1 [see Additional file 1]. Mean BMIs in 1987, 1999 and 2006 were 20.4, 20.7 and 21.4 for males (F = 17.0, p =.000), and 21.0, 21.5 and 22.1 for females (F = 15.9, p =.000). Male BMI increases over the 12 year period from 1987 to 1999 (F = 1.6, p =.206) were therefore somewhat smaller than those over the 7 year period from 1999 to 2006 (F = 24.7, p =.000). The prevalence of male obesity increased from 6.7% in 1987 to 10.6% in 1999 and 15.9% in 2006 (earlier increase p =.067; later increase p =.000). Corresponding rates for females were 5.4%, 11.5% and 14.9% (earlier p =.003; later p =.014). Thus, for males, the increase in prevalence of obesity, like BMI, was steeper over the 1999–2006 period. In addition (not shown on table), among obese males, mean (SD) BMIs also increased over time, from 26.2 (1.1) in 1987 to 27.0 (2.2) in 1999 and 28.3 (2.9) in 2006 (F = 12.4, p =.000). No such increase was seen among obese females, among whom mean (SD) BMIs were 28.4 (3.1), 28.7 (2.7) and 28.7 (2.8) in 1987, 1999 and 2006 (F = 0.1, p =.932).
Changes in SES reflect those in society over this period , with some minor variations in area deprivation. GHQ 'caseness' increased significantly among females over the earlier time period (p =.000) and among both males and females over the later one (p =.000 for both genders) (see also ). A clear gender difference in 'low self-esteem' can be seen at each date, but because this variable was defined as those with scores in the lowest quartile in each study, changes over time cannot be identified. Finally, the proportion reporting 'a lot' of weight worries also increased over time. The gender difference was greatest in 1987 (6.2% males worried 'a lot', compared with 29.0% females), and increases between 1987 and 1999 were greater for males, but significant for both genders (male p =.000; female p =.005). Thereafter, weight worries remained stable among males, but increased slightly among females (p =.005).
Table 2 [see Additional file 2] shows associations between obesity and both social class and area deprivation category, for males and females at each date. There were no significant social class differences in obesity at any date, for either males or females, and no differences according to area deprivation among males. However, among females, there were significant differences according to deprivation in 1987 (p =.000) and 2006 (p excluding 'missing' category =.035). At both dates, lowest prevalence occurred among those resident in less deprived areas, although at the later date the association was not linear.
Finally, the table shows that obesity increased between 1987 and 2006 for all gender and social class and area deprivation groups, apart from females from the most deprived areas.
Table 3 [see Additional file 3] shows that, apart from significantly increased odds of 'low' self-esteem among obese compared with non-obese females in 2006, neither GHQ-12 'caseness' nor 'low' self-esteem was associated with obesity at any date for either males or females (see also ).
At each date, obese males were around twice as likely as their non-obese peers to report weight worries, although wide CIs mean this was not significant in 1987. A trend towards a weakening relationship between obesity and male weight worries over time was not significant. In 1987 and 2006, but not 1999, obese females were, like males, around twice as likely as their non-obese peers to report weight worries. Despite the rather different association in 1999, the date by obesity interaction for female weight worries was not significant. At each date, considerable proportions of the obese did not report 'a lot' of worry about weight (e.g. 71.2% males, 40.2% females in 2006), while a significant minority of the non-obese did report 'a lot' of worry (e.g. 18.8% males, 40.1% females in 2006).