We found that the incidences of obesity and extreme obesity rose with increasing baseline BMI. The incidences of obesity and extreme obesity were higher among younger adults, women, non-Hispanic blacks and Hispanics, adults with lower levels of education, and adults who did not participate in any leisure-time physical activity. Adults who did not drink any alcoholic beverages had higher incidences than those who had at least one drink but were not heavy drinkers during the past month. Many, but not all, of these associations agree with the results of studies based on the prevalence of obesity. We also found that a considerable proportion of obese and extremely obese adults lost weight between 2008 and 2009 and moved to lower BMI categories. The mean weight loss increased as baseline BMI increased.
Baseline BMI was the most significant indicator for both obesity and extreme obesity incidence. This result was not surprising because adults with high baseline BMI would have had to gain less weight to become obese or extremely obese. This finding indicates that prevention of further weight gain should be the first step of obesity control, especially among "at-risk" adults with high baseline BMI, because obesity-related morbidities increase with increasing BMI [4, 5].
The incidences of obesity and extreme obesity were highest among adults aged 18 to 29 years, indicating that young adults are more likely to develop a weight problem even though the prevalence of obesity is lowest among this group . Obesity is associated with morbidity and with the leading causes of death in the United States [4, 5]. The risk for obesity-related chronic diseases will be significantly increased among young adults, and their quality of life will be considerably diminished throughout the rest of their lives once they become obese. Therefore, obesity prevention efforts are likely to have the biggest impact on young adults in their 20s. Consistent with previous studies [16, 17], we found that women were more likely to develop obesity than were men. These findings suggest that young adults, particularly young women, are important groups to focus on to prevent obesity.
Non-Hispanic blacks, Hispanics, and adults with lower levels of education had higher incidences of obesity and extreme obesity. Behavioral, cultural, and environmental factors may have contributed to the high incidences. According to one study, both non-Hispanic black women and Hispanic women are more satisfied with their body size than are non-Hispanic white women; those who are satisfied with their body size are less likely to try to lose weight . Evidence also suggests that black, Hispanic, and lower-income neighborhoods have fewer chain supermarkets and produce stores and less access to physical activity facilities; this limited access may negatively impact diet and physical activity levels .
Using the same data source, the 2009 BRFSS, a previous study indicated that the South and Midwest had higher prevalences of obesity than the Northeast and West . Our study shows that the South has a significantly higher incidence of extreme obesity than the Northeast, West, and the territories. The South may be a geographic region that warrants extra obesity prevention efforts.
Certain behavioral factors were associated with the incidences of obesity and extreme obesity even after controlling for baseline BMI and socio-demographic characteristics. Participating in any leisure-time physical activity was associated with decreased risks of developing obesity and extreme obesity. Physical activity plays a role in the maintenance of a healthy body weight, the loss of excess body weight, and the maintenance of successful weight loss because of its role in energy balance . Increasing physical activity among US adults through informational, behavioral, and environmental evidence-based approaches is important for obesity prevention .
As indicated in our study, any alcohol drinking was related to a decreased risk for obesity and extreme obesity compared to no alcohol drinking. This finding was similar to the results from a prospective cohort study conducted by Wang and colleagues . They concluded that normal weight middle-aged and older women who consumed a light to moderate amount of alcohol had a lower risk of becoming overweight and/or obese during 12.9 years of follow-up compared to nondrinkers. However, our finding should be interpreted with caution because our nondrinker group not only included those who never consume alcohol, but also former drinkers. The underlying mechanism for the association between obesity and alcohol consumption is complex and needs to be better understood. Studies found that some drinkers, especially female drinkers, tend to substitute alcohol for other foods without increasing total calorie intake, and lower intake of carbohydrates was related to higher levels of alcohol intake [21, 22].
Similar to a cohort study conducted by Watari and colleagues , we found that current and former smokers had a significantly higher incidence of obesity compared to nonsmokers. However, findings from other published studies that examined the relationships between smoking and BMI or prevalence of obesity have been inconsistent [12, 24]. Clarification of the mechanism that explains this association is of considerable interest.
Study strengths and limitations
The study's sample size, one of its strengths, was large enough to estimate incidence for subgroups and to ensure sufficient statistical power to detect differences across groups. Second, as the largest population-based telephone survey of adults in the United States, BRFSS allows us to obtain incidence estimates that represent all 50 states, the District of Columbia, and three US territories.
The findings in this report are subject to several limitations. First, our estimates are based on a cross-sectional survey rather than following people over time, and this limited our ability to distinguish people who were truly incident cases from those who had been obese in the past, but subsequently lost weight and then regained weight during the previous year (recurrent cases of obesity). We also assumed that the risk factors assessed in 2009 accurately reflected risk-factor status in 2009, and that these risk factors did not differ between incident and recurrent cases of obesity. Second, BMIs were based on reported weight and height, and it is widely known that these estimates, particularly among people with high BMIs, are underestimates [25, 26]. Previous studies have found, however, that recalled past weight is strongly correlated with measured weight and that self-reported weight change is reliable [27–29]. Although it is likely that biases in self-reported current and previous weights are correlated, this has not been documented, and our findings need to be confirmed by studies that include measured weights and heights. Third, the survey lacks complete dietary intake data, so we were not able to include all dietary behavioral factors or calorie intake in our modeling analyses. Fourth, the BRFSS excludes people who do not have landline telephones. Because adults who live in wireless-only households tend to be younger, male, Hispanic, binge drinkers, or current smokers, and have lower incomes , our incidence estimates may not be generalizable to the entire US population. Based on Council of American Survey and Research Organizations (CASRO) guidelines, the median response rate (percentage of all eligible people who completed interviews) in 2009 was only 52.5% (range: 37.9%- 66.9%), possibly resulting in biased estimates .