Our period by age and period by birth cohort analyses reveal the extent to which the epidemic of obesity has reached across age groups and birth cohorts in the US, and our age group by birth cohort analysis illustrates the dramatic shift in patterns of obesity prevalence over the life course that have occurred related to increases in obesity among younger individuals in the population. Because recent cohorts of individuals are reaching a higher prevalence of obesity earlier in the life course, they are experiencing a greater cumulative exposure to excess weight over their lifetime, which will likely have profound implications for future rates of obesity-related chronic diseases and mortality within the US population.
In 2003, Narayan et al used a Markov chain model to project the estimated lifetime risk and mean age at diagnosis of type 2 diabetes in the US for individuals born in the year 1984-2000.(11
) They estimated a lifetime risk of 32.8% and 38.5% of US males and females, with a mean age at diagnosis of type 2 diabetes ranging from 55.7-58.8 years depending on the age and gender group. Our results suggest that they underestimated future rates of type 2 diabetes, as they assumed constant age-specific rates of diabetes prevalence and incidence based on individuals from the 2000 National Health Interview Surveys, some of whom were from much older cohorts. For example, rates of diabetes among 50-59 year olds were based on individuals from the 1936 and 1926 cohorts, for whom obesity rates were substantially lower throughout the life course. The assumption by Narayan et al of constant rates therefore did not account for the increasing prevalence of obesity among younger individuals in the population.
Furthermore, the mean age of diagnosis of type 2 diabetes may be even younger than projected by Narayan et al, given that they did not account for the increasing duration of obesity over the life course occurring among younger cohorts. Independent of the severity of obesity, studies have shown that obesity duration is an important risk factor for type 2 diabetes,(15
) suggesting that a cumulative exposure to excess body weight may hasten the onset of type 2 diabetes among individuals. The increasing prevalence and duration of obesity occurring among the younger cohorts in the population may therefore lead to a shift in the epidemiology of type 2 diabetes within the population, with an earlier age of onset of type 2 diabetes during young or middle adulthood rather than later in adulthood. Such a shift will presumably have important consequences for future morbidity and mortality, independent of absolute increases in rates of type 2 diabetes, as an earlier age of onset of type 2 diabetes implies that individuals will have a longer duration of diabetes over their lifetime. This longer duration may accelerate rates of complications such as cardiovascular disease, retinopathy, and end stage renal disease within the population,(17
) with potential implications for estimates of future life expectancy in the US.
In 2005, Olshansky et al(8
) made the provocative prediction that the epidemic of obesity would result in potential declines in US life expectancy in the 21st
century. Using a life table approach, they calculated the reduction in death rates that would occur if all obese individuals in the US population had an optimal adult body weight (BMI=24), forecasting reductions in life expectancy of anywhere from one third to three fourths of a year. However, their assumptions about obesity prevalence were based on estimates from the period 1988-1994,(8
) which represented just the midpoint of the obesity epidemic, and were based only on adults, the majority of whom were from much older birth cohorts (1885 to 1976). Therefore, their predictions likely represent an underestimate of reductions in life expectancy related to healthy versus obese BMI. The greater prevalence and duration of obesity seen among younger cohorts compared with older cohorts warrant further studies to better understand the future impact of these birth cohort trends on the age of onset and incidence of type 2 diabetes, other obesity-related chronic diseases, and overall life expectancy in the US population.
The Centers for Disease Control (CDC) recently reported no significant changes in obesity prevalence between 2003-2004 and 2005-2006 for both US children and adults,(19
) leading some to suggest that recent public health efforts may have been effective in “halting” further development of the epidemic. Their findings are consistent with our analyses at least for the older birth cohorts, given that we observed an attenuation of absolute percentage increases in obesity prevalence leading to the most recent decade of life compared with the previous decade of life. However, this attenuation was not seen for the more recent 1976-85 birth cohort; in fact there was an acceleration of obesity trends, between the transition from adolescence (10-19 years) to young adulthood (20-29 years). It has been hypothesized that adolescence is a critical period for development of obesity, which may be associated with greater risk for persistence of obesity and obesity-associated morbidity in adulthood.(21
) Therefore, we speculate that long-term trends in obesity may be different for more recent cohorts.
However, even under the assumption that period trends in obesity are stabilizing across most birth cohorts, our life course approach places these trends in a different perspective. Even if the plateau in obesity prevalence persists beyond this 2-year period trend, because younger generations already are carrying a higher obesity prevalence as well as a longer duration of obesity for a given age compared with previous generations, this raises the question of whether “halting” or stabilizing the epidemic currently will be enough to change future rates of obesity-related morbidity and mortality among younger generations. Our findings suggest that initiatives must not only halt but actually reverse population trends in obesity in order to return risks of obesity-related complications to levels experienced by older generations.
The differences in age group by birth cohort obesity trends by race and gender have important implications for disparities in chronic disease over the life course. Improvements in obesity trends leading to the most recent decade of life were greater for white compared with black cohorts. Given that absolute obesity prevalences are already higher for black cohorts compared with white cohorts, such a trend may lead to further increases in the disproportionate burden of chronic disease among the black population in the US.(22
Regarding gender differences, improvements in obesity trends leading to the most recent decade of life were greater for female compared with male cohorts. Therefore, although females have higher absolute obesity prevalence compared with males, US males may “catch-up” with their female same-age peers with regard to obesity prevalence. These recent race- and gender-specific trends will be critical for future projection of chronic disease rates and life expectancy.
Our study is novel, in that it evaluated age, period, and cohort trends in US obesity prevalence over the life course, from 2 through 79 years of age. We also looked at cohort trends in BMI, which were similar to the obesity trends, indicating that BMI distribution has shifted across the population, regardless of weight status. Wang et al recently performed a systematic review and meta-analysis of studies published on the obesity epidemic in the US, but only starting from 1990, and evaluated just period estimates of obesity prevalence, separately for children and adults.(23
) Other studies have evaluated cohort trends in obesity, but have either exclusively focused on the elderly, with the 1956-1965 birth cohort as the youngest cohort of interest,(24
) or predicted obesity rates for older adults.(25
) Therefore, this is the first study to account for the potential effects of the childhood obesity epidemic on birth cohort differences in obesity over time. Other strengths include the use of nationally representative data, the availability of measured height and weight, and the ability to evaluate trends separately by gender and race.
Although we would like to evaluate birth cohort trends in obesity related complications, focusing in particular on the youngest US individuals, the current prevalence estimates of chronic disease in younger individuals are quite low; for example, the estimated prevalence of type 2 diabetes among US children and adolescents (0-19 years) is only 0.02%.(26
) As a result, assessing these trends for younger individuals using NHANES would likely result in unreliable estimates. This is consistent with our hypothesis that the increasing duration and prevalence of obesity in the childhood population will not have implications for development of chronic obesity-related diseases until decades later in life, as these children age into adulthood.(27
We acknowledge that there are limitations to our study. The trends we identified must be cautiously interpreted given that the birth cohorts that we created did not consist of the same individuals across all waves. For example, it is possible that obesity prevalence was underestimated among the older age groups (60-69 years and 70-79 years), given the known association between obesity and excess mortality.(5
) Although age period cohort analyses are ideally conducted using longitudinal data, to our knowledge no such data exist over a similar time span for a representative sample of the US population.
Our assumptions about the year of the interview could lead to misclassification of individuals into birth cohorts. However, the purpose of our study was to identify overall population trends, and these trends were consistent in our sensitivity analyses (data not shown). We recognize that trends in obesity differ dramatically by gender and race; for example, prevalence estimates of obesity among black females are significantly higher compared with black males and white females.(23
) Unfortunately, we were unable to evaluate age by birth cohort trends for gender- and
race-specific groups, due to smaller sample sizes resulting in unstable estimates. We were unable to generate analyses for Mexican-Americans due to the fact that they were not oversampled until the later NHANES surveys; other ethnic groups (Asian-Americans and other Hispanic) could not be included since they are not oversampled. Although individuals were classified as either black or white for NHANES 1971-1975 and 1976-1980, they were classified as non-Hispanic white and non-Hispanic black for the remaining surveys.
Finally, BMI is an inexpensive method for assessing body fat in large epidemiologic studies, but BMI has its limitations as a measure. One study showed that for a comparable BMI, older persons have higher body fat compared with younger persons, and women have higher body fat compared with men; therefore we may have underestimated obesity among these groups.(28