Our data demonstrate that the obesity epidemic continued in this fairly representative, almost exclusively Caucasian population in Norway. It is worrying that the greatest increase was seen in the younger adult groups, as the risk of obesity complications increases with duration of obesity . An important finding is, however, that the increase in overweight was significantly smaller in the second period (between HUNT2 and HUNT3); 1.9% (from 50.5 to 52.4% in men) and 0.6% correspondingly in women compared with 8.4% in men and 7.2% in women, respectively, in the first period (between HUNT1 and HUNT2). This considerably smaller increase in overweight may give hope of a future reduction in the increase of obesity in Norway in line with studies indicating that the obesity epidemic is already slowing down [3-5], in the United States most pronounced in women , but recently also in men .
Nevertheless, we observed a shift in the distribution curves of BMI and WC to the right, indicating that the change was not only due to fat people getting fatter. The shift in BMI was considerably greater than that observed for American adults between 1999/2000 and 2007/2008 . Similar patterns were observed in 40–42-year-old Norwegians from three other counties . Additionally, both self-reported  and measured  data demonstrated that overweight and obesity also increased in children in Norway. In addition to this general shift towards higher body weight in all weight categories, there was an especially large increase in obesity classes II and III even between HUNT2 and HUNT3 in most age groups of men and younger age groups of women.
In the US National Health and Nutrition Examination Surveys , which also used measured data, the prevalence of obesity in non-Hispanic white men aged ≥20 years increased from 27.3% in 1999–2000 to 33.1% in 2005–2006 and 31.9% in 2007–2008. In our study, the obesity prevalence in the same age and sex group increased from 14% in 1995–1997 to 22.1% in 2006–2008. Although our study showed lower obesity prevalence than the US comparable group, the sum of overweight and obesity groups in the latest surveys (BMI ≥ 25 kg m−2) was similar (HUNT 67.0% vs. the United States 67.5%). In a recent worldwide review, Finucane et al.  estimated a global increase in mean BMI of 0.4 kg m−2 per decade. The corresponding HUNT figures for the 22-year period covered was 1.0 kg m−2 per decade compared with 1.1 kg m−2 per decade for the United States.
BMI is the most widely applied measure in epidemiological studies, but it has well-known limitations as a surrogate measure for body fat, especially in those with high muscle mass. Furthermore, it has recently been shown that many people identified with obesity based on body fat measurements had BMI in the normal range . WC, reflecting central or visceral obesity, has gained increasing attention. Both BMI and WC are independently associated with cardiovascular risk and risk of death , but their relative importance may differ depending on the end point  or sex . Future health costs may be predicted better by WC than by BMI , and WC may precede other cardiovascular risk factors . The distribution curves of BMI and WC were comparable in our study. Obesity was more prevalent and increased to a greater extent when defined by WC than by BMI, indicating a considerable increase in abdominal fat. WC is a strong and additional risk factor for type 2 diabetes and all-cause mortality , and is more strongly related to mortality than BMI in persons with diabetes , raising health-related concerns in our population. The reason for this distinct increase in WC is difficult to establish. Compared with HUNT2, the proportion of HUNT participants reporting at least 30 min of daily physical activity was greater for HUNT3 in all age groups (data not shown). The amount of sitting time has, however, increased between HUNT2 and HUNT3. Sitting time is positively associated with cardiovascular risk factors (Chau J et al. personal communication), and one could speculate that sitting time might affect WC more than BMI. Further research is needed to examine this against more precise measures of body and abdominal fat.
Although BMI-defined obesity prevalence increased more in men than in women between HUNT2 and HUNT3, the reverse was observed for WC-defined obesity prevalence. Mean WC increased by 8.9 cm in women and 6.5 cm in men over the same period. A similar increase was observed in both sexes in Canada (10.6 and 6.5 cm, respectively) , but over a longer period (1981–2007/2009). A greater increase in women is also reported in Finland . In the United States, a much smaller increase, with similar increases between the sexes (2.9 cm in men and 3.2 cm in women), was observed between 1988/1994 and 1999/2000 .
The same gender differences were found in a recent study from Young-HUNT (the adolescent part of the HUNT Study), following changes in BMI and WC in 13–19-year-old participants with normal weight prospectively for 11 years . While 8% of girls and 9% of boys developed BMI-defined obesity, 34% of girls and 9% of boys developed WC-defined obesity.
WC is reported to predict all-cause mortality better in women than in men , which make the observed change in WC in women a worry. The substantial increase in WC in women could, at least partly, be related to changes in androgen levels in women [47, 48] because total fat mass in women correlates with circulating testosterone levels . It is well known that abdominal obesity increases among women after menopause, associated with an increased amount of bioavailable testosterone , although this would not explain the increase in pre-menopausal women. Testosterone levels were not measured in our study.
Strengths and limitations
In contrast to most other studies, the present study includes a large population with wide age span and long follow-up period. Objective standardised measurements of height and weight were performed by trained personnel, and are more reliable than self-reported data used in many studies. Self-report has been shown to overestimate height and underestimate weight (especially in women), both resulting in lower than actual BMI [51, 52], although the degree of underestimation varies .
A limitation of the study might be the declining participation rate from HUNT1 to HUNT3, introducing a potential selection bias, especially in the younger age groups. Overweight people might be more inclined to participate in the later surveys, but the opposite could also occur because of fear of unwanted comments on their body weight. Non-participation studies in HUNT1 and HUNT2 demonstrated that young people mostly attributed non-attendance to forgetfulness or practical difficulties like being away for school or work. After HUNT3, a two-page questionnaire was sent to all non-attendees and 6923 persons (16%) answered. BMI based on self-reported height and weight in this group was, in all age and sex groups, slightly lower than the measured values in those who attended . The mean difference in BMI (0.6 kg m−2 in men and 1.1 kg m−2 in women) was similar in men, but higher in women to that reported in an Australian study, also performed in 2007–2008 (0.6 vs. 0.7 kg m−2 in men and women, respectively) . This might imply that the high prevalence of obesity in men in HUNT3 is not due to the lower attendance, particularly in men, observed in HUNT3.
As the Nord-Trøndelag County is mainly rural with mean income and education levels slightly lower than the national average, the comparability to the Norwegian population might be attenuated. The obesity prevalence in the city of Tromsø in 1994–1995 was 9.5% in men and 11.5% in women compared with 14.4 and 18.3%, respectively, in those of similar age group in HUNT2 (1995–1997) . Accordingly, less densely populated areas in Nord-Trøndelag County had more obesity than the county's towns (data not shown). Because overweight is inversely related to socioeconomic status , this might also have contributed to some overestimation of obesity prevalence in the present study. However, the increasing non-western immigrant population in the major cities, acknowledged to be more prone to obesity, might have led to an underestimation. Obesity in certain age groups in Tromsø nearly doubled, especially in men, from 1994–1995 to 2001 . Altogether, we have no evidence that these factors in Nord-Trøndelag have changed significantly over the period of time studied. We, therefore, consider that the observed changes in this study are generally relevant to the Norwegian population.
A third potential limitation might be the 11-year span between the surveys; therefore, we cannot be certain about what has happened in-between. Other Norwegian studies indicate that there has been a continuous increase in body weight in both 40–42-year-olds [11, 33] and children . However, we cannot completely rule out an initial increase between HUNT2 and HUNT3 followed by a plateau in weight increase.
Our study showed that the increase in obesity in this relatively representative Norwegian adult population continued to increase up to 2008. Of particular concern is the considerable increase in young adults. The increase in BMI-defined obesity was greater in men than in women. The increase in abdominal obesity was even greater and present in all age and sex groups, but the increase was greater in women than in men for almost all age groups. A slower increase in the overweight group may indicate a future plateuing of the increase in obesity. Because of the increase in obesity in this Caucasian population, there is concern that the decreasing cardiovascular morbidity and mortality observed in Norway may reverse when the full effect of smoking cessation is achieved, unmasking consequences of the increasing obesity. More extensive and effective prevention strategies are recommended to avoid a future reduction in the public's health.