Elsie R. Pamuk

The Effect of Age on the Association between Body-Mass Index and Mortality

BACKGROUND The effect of age on optimal body weight is controversial, and few studies have had adequate numbers of subjects to analyze mortality as a function of body-mass index across age groups. METHODS We studied mortality over 12 years among white men and women who participated in the American Cancer Societys Cancer Prevention Study I (from 1960 through 1972). The 62,116 men and 262,019 women included in this analysis had never smoked cigarettes, had no history of heart disease, stroke, or cancer (other than skin cancer) at base line in 1959-1960, and had no history of recent unintentional weight loss. The date and cause of death for subjects who died were determined from death certificates. The associations between body-mass index (defined as the weight in kilograms divided by the square of the height in meters) and mortality were examined for six age groups in analyses in which we adjusted for age, educational level, physical activity, and alcohol consumption. RESULTS Greater body-mass index was associated with higher mortality from all causes and from cardiovascular disease in men and women up to 75 years of age. However, the relative risk associated with greater body-mass index declined with age. For example, for mortality from cardiovascular disease, the relative risk associated with an increment of 1 in the body-mass index was 1.10 (95 percent confidence interval, 1.04 to 1.16) for 30-to-44-year-old men and 1.03 (95 percent confidence interval, 1.02 to 1.05) for 65-to-74-year-old men. For women, the corresponding relative risk estimates were 1.08 (95 percent confidence interval, 1.05 to 1.11) and 1.02 (95 percent confidence interval, 1.02 to 1.03). CONCLUSIONS Excess body weight increases the risk of death from any cause and from cardiovascular disease in adults between 30 and 74 years of age. The relative risk associated with greater body weight is higher among younger subjects.

Inequality in income and mortality in the United States: analysis of mortality and potential pathways

Abstract Objective: To examine the relation between health outcomes and the equality with which income is distributed in the United States. Design: The degree of income inequality, defined as the percentage of total household income received by the less well off 50% of households, and changes in income inequality were calculated for the 50 states in 1980 and 1990. These measures were then examined in relation to all cause mortality adjusted for age for each state, age specific deaths, changes in mortalities, and other health outcomes and potential pathways for 1980, 1990, and 1989-91. Main outcome measure: Age adjusted mortality from all causes. Results: There was a significant correlation (r=0.62, P<0.001) between the percentage of total household income received by the less well off 50% in each state and all cause mortality, unaffected by adjustment for state median incomes. Income inequality was also significantly associated with age specific mortalities and rates of low birth weight, homicide, violent crime, work disability, expenditures on medical care and police protection, smoking, and sedentary activity. Rates of unemployment, imprisonment, recipients of income assistance and food stamps, lack of medical insurance, and educational outcomes were also worse as income inequality increased. Income inequality was also associated with mortality trends, and there was a suggestion of an impact of inequality trends on mortality trends. Conclusions: Variations between states in the inequality of the distribution of income are significantly associated with variations between states in a large number of health outcomes and social indicators and with mortality trends. These differences parallel relative investments in human and social capital. Economic policies that influence income and wealth inequality may have an important impact on the health of countries. Key messages There was a significant correlation (r=0.62) between the proportion of total household income received by the less well off 50% of households and variation between states in death rates for the United States Income inequality was also significantly related to changes in mortality with smaller declines between 1980-90 in those states with greater income inequality Income inequality was associated with a large number of other health outcomes and with measures related to investments in human and social capital Economic policies that increase income inequality may also have a deleterious effect on population health

Income inequality and mortality in metropolitan areas of the United States.

OBJECTIVES This study examined associations between income inequality and mortality in 282 US metropolitan areas. METHODS Income inequality measures were calculated from the 1990 US Census. Mortality was calculated from National Center for Health Statistics data and modeled with weighted linear regressions of the log age-adjusted rate. RESULTS Excess mortality between metropolitan areas with high and low income inequality ranged from 64.7 to 95.8 deaths per 100,000 depending on the inequality measure. In age-specific analyses, income inequality was most evident for infant mortality and for mortality between ages 15 and 64. CONCLUSIONS Higher income inequality is associated with increased mortality at all per capita income levels. Areas with high income inequality and low average income had excess mortality of 139.8 deaths per 100,000 compared with areas with low inequality and high income. The magnitude of this mortality difference is comparable to the combined loss of life from lung cancer, diabetes, motor vehicle crashes, human immunodeficiency virus (HIV) infection, suicide, and homicide in 1995. Given the mortality burden associated with income inequality, public and private sector initiatives to reduce economic inequalities should be a high priority.

The Importance of Place of Residence: Examining Health in Rural and Nonrural Areas

We examined differences in health measures among rural, suburban, and urban residents and factors that contribute to these differences. Whereas differences between rural and urban residents were observed for some health measures, a consistent rural-to-urban gradient was not always found. Often, the most rural and the most urban areas were found to be disadvantaged compared with suburban areas. If health disparities are to be successfully addressed, the relationship between place of residence and health must be understood.

Weight Control Practices of U.S. Adolescents and Adults

Weight control is an important concern of adolescents and adults for reasons of both health and physical appearance. Although weight control may decrease the risk for chronic disease in adulthood, overemphasis on thinness in adolescence can lead to unhealthy weight-loss practices and may contribute to the development of eating disorders [1]. Although attempts at weight control are common in the U.S. population, little current information on the prevalence of weight control practices is available from population-based studies. Most previous surveys of weight control practices have been conducted among students attending a single high school or college or among persons enrolled in weight control programs. This report estimates the prevalence of various body weight perceptions and weight control practices among U.S. adolescents and adults. We examined data from two surveys: the Youth Risk Behavior Surveillance System, which collected data from a representative sample of U.S. high school students in 1990, and the Behavioral Risk Factor Surveillance System, which collected data from adults in 38 states and the District of Columbia in 1989. Methods Adolescents The national school-based Youth Risk Behavior Survey is used to monitor the prevalence of health risk behaviors among high school students [2, 3]. The 1990 national school-based Youth Risk Behavioral Survey used a self-administered, 75-item questionnaire given within one class period during the spring. Students from public and private schools were selected using a three-stage cluster design to obtain a sample of students in grades 9 through 12 that was representative of the 50 states, the District of Columbia, and the Virgin Islands. Schools with a high proportion of black and Hispanic students were oversampled. The response rate for schools was 74% (124 of 168 schools), and that for students was 87% (11 631 of 13 402 students). We excluded 164 students because of inadequate information on demographics or weight control practices. The final sample included was 11 467 students. Our analysis weighted all responses to compensate for the sampling design. Respondents were asked the following questions about weight perception and weight control practices: 1) Do you think of yourself as too thin (underweight), about the right weight, or too fat (overweight)? 2) Which of the following are you doing about your weight? Not trying to do anything about weight, trying to lose weight, trying to keep from gaining more weight, or trying to gain more weight? 3) During the past 7 days, how many meals did you skip to try to lose weight or to keep from gaining weight? None, 1 or 2 meals, 3 to 6 meals, 7 to 14 meals, 15 or more meals. 4) During the past 7 days, how many times did you take a diet pill, vomit on purpose, or exercise to try to lose weight or keep from gaining weight? Never done this; have done this but not in the past 7 days, 1 or 2 times, 3 to 6 times, 7 to 14 times, 15 or more times. For the purpose of analysis, we defined ever use of these specific weight control methods to include one or more times in the past 7 days or have done this but not in the past 7 days. Adults Data for the Behavioral Risk Factor Surveillance System were collected by state health departments in collaboration with the Centers for Disease Control. The primary purpose of this system was to provide state-specific estimates of behaviors that relate to the leading causes of death among U.S. adults. In each participating state, an independent probability sample of adult residents with telephones was selected using a multistage cluster sampling design based on the Waksberg method [4]. All states used an identical core questionnaire administered by trained interviewers. Interviews lasted approximately 25 minutes [5]. During the interview, all respondents were asked, Are you now trying to lose weight? Those who answered no were asked, Are you now trying to maintain your weight; that is, keep from gaining weight? After the questions on weight control practices, all respondents were asked, Do you now consider yourself to be overweight, underweight, or about average? Questions about self-reported weight and height were asked at the end of the interview [6]. During 1989, 38 states and the District of Columbia collected data. One state was excluded from analysis because information was not collected on weight control practices. We limited our analysis to the 64 311 persons who were not pregnant at the time of the interview. We excluded an additional 3 450 persons because of inadequate information on sociodemographic status, weight perception, weight control goals, or current weight and height. Our final sample included 60 861 participants. The median response rate for the states was 82%. We directly age-standardized all prevalence estimates using the age distribution of the total Behavioral Risk Factor Surveillance System sample as the referent population. We categorized respondents according to three categories of body mass index based on the National Health and Examination Survey (NHANES) II reference sample for persons 20 to 29 years old (< 85th percentile, 85th to < 95th percentile, and 95th percentile) [7]. Results Adolescents Among All Students The Youth Risk Behavior Survey found substantial differences in weight perception between male and female students. Female students were more than twice as likely as male students to consider themselves to be too fat (34% and 15%, respectively) and were less than half as likely to consider themselves to be too thin (7% and 16%, respectively). Among female students, 37% of both whites and Hispanics considered themselves to be too fat compared with 25% among blacks. Among male students, whites and Hispanics were also more likely than blacks to consider themselves to be overweight (16%, 15%, and 8%, respectively). Among female students, 44% reported that they were trying to lose weight, 26% were trying to keep from gaining weight, 7% were trying to gain weight, and 23% were not trying to do anything about their weight (Table 1). Among male students, 15% reported that they were trying to lose weight, 15% were trying to keep from gaining weight, 26% were trying to gain weight, and 44% were not trying to do anything about their weight. Among both male and female students, attempts to lose weight were unrelated to grade level and were most strongly associated with weight perception, although less so among blacks. The students reported using the following methods to lose or maintain weight in the 7 days preceding the survey: exercising (51% of female students and 30% of male students), skipping meals (49% and 18%, respectively), using diet pills (4% and 2%, respectively), and vomiting (3% and 1%, respectively). In general, the percentage of students who reported ever having used these methods was much higher: exercising (80% and 44%, respectively), using diet pills (21% and 5%, respectively), and vomiting (14% and 4%, respectively). Table 1. Prevalence of Current Weight Control Practices among High School Students, by Selected Characteristics from the Youth Risk Behavior Survey, 1990 Among Students Trying To Lose Weight Among both sexes, exercise and skipping meals were much more frequently used to lose weight than were either diet pills or vomiting (Table 2). Exercising once a day or more as a means of weight control was more frequent among male students (29% of male students compared with 19% of female students), whereas skipping meals at least daily was more common among female students (17% of female students compared with 11% of male students). Table 2. Prevalence and Reported Frequency of Specific Weight Control Practices among High School Students Reported in the Past 7 Days among Those Who Are Currently Trying to Lose Weight by Selected Characteristics from the Youth Risk Behavior Survey, 1991* Compared with female students trying to lose weight, male students were slightly more likely to use exercise and less likely to skip meals. Among female students trying to lose weight, use of exercise in the previous week was somewhat higher among those in their first 2 years of high school (see Table 2). Use of other methods did not vary consistently by grade level. White students were most likely and black students were least likely to use exercise for weight control. Among male students who were trying to lose weight, use of diet pills and exercise increased with increasing grade level; however, use of other methods did not vary consistently by grade. Compared with white students, black students were somewhat less likely to report exercise as a means of weight control. Adults Women were more likely than men to consider themselves to be overweight (38% compared with 28%). White, black, and Hispanic women were equally likely to consider themselves overweight. Among men, whites and Hispanics were somewhat more likely than blacks to consider themselves overweight (29%, 28%, and 23%, respectively). Among women, 38% reported that they were trying to lose weight at the time of the survey, 28% were trying to maintain weight, and 33% were doing neither. Among men, 24% reported that they were trying to lose weight, 28% were trying to maintain weight, and 48% were doing neither. Among both sexes, the prevalence of attempts at weight loss was similar among all participants up to 59 years old but decreased substantially at older ages (Table 3). Among women, the prevalence of attempts at weight loss was similar across ethnic and education groups. Among men, Hispanics and the more educated were more likely to report trying to lose weight. Among both sexes, the practice of trying to lose weight was strongly associated with body mass index and weight perception. Table 3. Prevalence of Current Weight Control Practices among Adults by Selected Characteristics from the Behavioral Risk Factor Surveillance System, 1989* Discussion Attempts at weight control are very prevalent among bo

Weight Loss and Subsequent Death in a Cohort of U.S. Adults

To date, observational studies have provided conflicting evidence about the relation between weight loss and death [114]. Much of the difficulty in interpreting these findings results from the inability to distinguish directly voluntary weight loss from that produced by illness. The authors of these studies have tried to control for the effect of illness-associated weight loss by some combination of statistical techniques to adjust for preexisting illness within the study population; exclusion of persons with known illnesses; or exclusion of early deaththose occurring within the first 2 to 5 years after the assessment of weight change. Reservations remain, however, regarding the adequacy of these approaches. Persons with some types of illness may lose weight voluntarily and may be advised by their physician to do so. Also, exclusion of deaths occurring within a few years of baseline may not account for all weight loss due to occult illness. An additional problem is that most studies have not been able to control for the potentially confounding effect of cigarette smoking [15]. Previous Analysis of Weight Loss and Death In a recently published study [16], we examined the relation between weight loss and subsequent death among 2140 men and 2550 women between 45 to 74 years old who participated in the first National Health and Nutrition Examination Survey (NHANES I, 1971 to 1975). Weight loss was assessed as the difference between self-reported maximum lifetime weight and weight measured at baseline. Vital status was determined through 1987. Risk for death associated with relative weight loss (< 5%, 5% to 14%, 15% or more) was estimated for three strata of maximum body mass index (BMI; weight[kg]/height[m]2)less than 26, between 26 and 29, and 29 or more. We used Cox proportional-hazards models to adjust for age, race, parity (for women), cigarette smoking (never, former, or current), and maximum BMI as well as preexisting illnesses associated with weight loss (cancer, chronic bronchitis, or emphysema), preexisting illnesses associated with excess weight (heart attack, heart failure, stroke, hypertension, and diabetes), and diagnosis of both types of conditions. In addition, the analysis excluded persons who died in the first 5 years after the baseline examination. We found that, at maximum BMIs of 29 or more, weight losses of 5% to 14% appeared to be protective for men but not for women. Among men and women whose maximum BMI was less than 29, the risk for death increased with increasing weight loss; among participants who had been moderately overweight (maximum BMIs between 26 and 29), those who lost 15% or more had more than twice the mortality risk of those losing less than 5%. Additional Analyses This report used several more rigorous exclusionary criteria to test the consistency of these findings. We first examined the effect of excluding early death by estimating relative risks associated with weight loss for the full cohort (2453 men and 2739 women), and for persons surviving for 5 and 8 years after the baseline examination, respectively. Within each stratum of maximum BMI, adjusted risk estimates for death from all causes, cardiovascular disease (ICD 9 codes 401 to 459 and 798), and noncardiovascular disease (all other ICD 9 codes except those for external injury) [17] were calculated relative to persons who had lost less than 5% of their maximum weight. We compared these results with those obtained by limiting the analysis to persons with no preexisting medical conditions except for hypertension or diabetes. Because weight loss may have implications for the elderly that are different than those for younger persons [9, 10], we also estimated adjusted mortality risks for persons who were 45 to 64 years old at baseline. The variables for smoking status included in the model may not have allowed adequate adjustment for confounding. We repeated the analysis of all-cause mortality risk for persons who, at baseline, reported that they had never smoked. Results The mortality risks associated with weight loss in men are shown in Table 1. If loss from maximum weight were a marker for occult illness, we would expect the relative risk estimates associated with weight loss to be reduced after excluding persons who died within the first 5 years after baseline and perhaps to be reduced even further after excluding deaths occurring in the first 8 years. Among men in this study, however, extension of the exclusionary period did not produce this type of consistent effect on the relative risk estimates. Table 1. Relative Risk for Death among Men 45 to 74 Years Old at Baseline, by Maximum Body Mass Index and Percentage of Maximum Weight Lost* After excluding deaths occurring in the first 8 years after baseline, the risk for death from noncardiovascular disease increased with the amount of weight lost among men with maximum BMIs of less than 26. Among men whose maximum BMI was between 26 and 29, weight losses of 5% to 14% were associated with an increased risk for death from cardiovascular disease, and weight loss of 15% or more was associated with a more than twofold mortality risk for both cardiovascular and noncardiovascular diseases. For men whose maximum BMI was 29 or more, however, weight losses of 5% to 14% appeared to reduce the risk for death from cardiovascular disease by approximately 30%. The risk for death from cardiovascular disease was not reduced among men who lost 15% or more, and the risk for death from noncardiovascular disease appeared to be moderately increased; however, this result was not statistically significant. The results for women are shown in Table 2. The exclusion of early death had little effect on the risk for death from cardiovascular disease. After limiting the analysis to women who survived at least 8 years after baseline, we continued to find a strong, direct association between weight loss and risk for death from cardiovascular disease among women whose maximum BMI was less than 29. We found a moderate, but not statistically significant, association for those with a maximum BMI of 29 or more. Extension of the exclusionary period, however, reduced the association between weight loss and death from noncardiovascular disease for women in the lowest and highest strata of maximum BMI. Table 2. Relative Risk for Death among Women 45 to 74 Years Old at Baseline, by Maximum Body Mass Index and Percentage of Maximum Weight Lost* We repeated the analysis after excluding all persons with previously diagnosed medical conditions except for hypertension and diabetes and after excluding deaths that occurred within the first 5 years after baseline. For both sexes, the relative risk estimates were similar to those obtained by excluding persons who died within the first 8 years (data not shown). The estimated relative risks for persons 45 to 64 years old at baseline were either the same as or higher than those estimated for the entire group (data not shown). Figures 1 and 2 show the relative risks for death due to all causes for participants who had never smoked at baseline. Because fewer than one third of the men in the study cohort had never smoked, we preserved sample size by excluding only men who died in the first 5 years after baseline. The results are consistent with those in Table 1, although the relative risk is not significantly elevated for men with maximum BMIs between 26 and 29 who lost 5% to 14% of their maximum weight (see Figure 1). Relative risk estimates for women who never smoked excluded deaths that occurred within the first 8 years after baseline. Compared with the results in Table 2, the relative risk for death associated with weight loss was only slightly lower for women who never smoked and only if their maximum BMI was less than 29. Figure 1. Relative risk for death among men who never smoked, by maximum body mass index and percentage of maximum weight lost. Figure 2. Relative risk for death among women who never smoked, by maximum body mass index and percentage of maximum weight lost. Discussion Observational studies have produced inconsistent results regarding the association between weight loss and risk for death. These studies have differed substantially in design and definition of weight loss and have not directly assessed the reason for the weight loss. Studies that have found an elevated risk for death among persons who have lost weight have generally assumed the finding to result from involuntary weight loss associated with illness [5, 6, 810]. An adverse effect of weight loss has also been found, however, in studies that excluded participants with known medical conditions [11, 12, 14]. Because weight loss may disproportionately benefit persons with conditions such as coronary heart disease, hypertension, or diabetes, exclusion of these participants may have obscured a positive effect of weight loss on death. The exclusion of participants with known medical conditions may not fully control for the effect of illness-associated weight loss on the risk for death because some illnesses may not have been diagnosed. The procedure usually used to remove the effect of subclinical illness involves the exclusion of death occurring within 2 to 5 years after baseline [15]. We showed that, among women, extension of the exclusionary period consistently attenuated the adverse association between weight loss and risk for death from noncardiovascular disease. This effect was strongest for women in the lowest and highest strata of maximum BMI. For both sexes, the relative risk estimates for persons who survived for at least 8 years after baseline were similar to those estimated after exclusion of persons with diagnosed illnesses other than hypertension and diabetes in addition to exclusion of those who died in the first 5 years after baseline. It therefore seems plausible that this attenuation resulted from removal of the influence of illness-associated weight loss. We should also consider

Estimating Deaths Attributable to Obesity in the United States

Estimates of deaths attributable to obesity in the United States rely on estimates from epidemiological cohorts of the relative risk of mortality associated with obesity. However, these relative risk estimates are not necessarily appropriate for the total US population, in part because of exclusions to control for baseline health status and exclusion or underrepresentation of older adults. Most deaths occur among older adults; estimates of deaths attributable to obesity can vary widely depending on the assumptions about the relative risks of mortality associated with obesity among the elderly. Thus, it may be difficult to estimate deaths attributable to obesity with adequate accuracy and precision. We urge efforts to improve the data and methods for estimating this statistic.

Are geographic regions with high income inequality associated with risk of abdominal weight gain

Geographic regions characterized by income inequality are associated with adverse mortality statistics, but the pathophysiologic mechanisms that mediate this ecologic relationship have not been elucidated. This study used a United States mail survey of 34158 male and 42741 female healthy-adult volunteers to test the association between residence in geographic regions with relative income inequality and the likelihood of weight gain at the waist. Respondents came from 21 states that were characterized by the household income inequality (HII) index, a measure reflecting the proportion of total income received by the more well off 50% of households in the state. The main outcome measure was self-reported weight gain mainly at the waist as opposed to weight gain at other anatomic sites. After controlling for age, other individual-level factors, and each states median household income, mens likelihood of weight gain at the waist was positively associated (p = 0.0008) with the HII index. Men from states with a high HII (households above the median receive 81.6% to 82.6% of the income) described weight gain at the waist more often than men from states with a low HII (households above the median receive 77.0% to 78.5% of the income) (odds ratio = 1.12, 95% confidence interval 1.03 to 1.22). Womens results showed a non-significant trend in the same direction. An association between ecologically defined socio-environmental stress and abdominal obesity may help to clarify the pathophysiologic pathways leading to several major chronic diseases.

Achieving National Health Objectives: The Impact on Life Expectancy and on Healthy Life Expectancy

Our study quantifies the impact of achieving specific Healthy People 2010 targets and of eliminating racial/ethnic health disparities on summary measures of health. We used life table methods to calculate gains in life expectancy and healthy life expectancy that would result from achievement of Healthy People 2010 objectives or of current mortality rates in the Asian/Pacific Islander (API) population. Attainment of Healthy People 2010 mortality targets would increase life expectancy by 2.8 years, and reduction of population wide mortality rates to current API rates would add 4.1 years. Healthy life expectancy would increase by 5.8 years if Healthy People 2010 mortality and assumed morbidity targets were attained and by 8.1 years if API mortality and activity limitation rates were attained. Achievement of specific Healthy People 2010 targets would produce significant increases in longevity and health, and elimination of racial/ethnic health disparities could result in even larger gains.

FLEGAL ET AL. RESPOND

Hu et al. speculate that the number of deaths attributable to obesity in the United States may be underestimated when relative risks are calculated on the basis of current body mass index (BMI). They cite no data to support their speculations, but instead invoke the notion of “reverse causality.” They hypothesize that relative risks are lowered by obese people who become ill, lose weight because of this illness to become normal weight, and die shortly thereafter of the underlying illness, surviving just long enough to be included in the study. However, this reverse-causation hypothesis is unlikely to be the correct explanation for the lower relative risks in the elderly for 2 reasons: first because data show that exclusion of preexisting illness has little effect on relative risk estimates,1 and second because weight loss from obesity to normal weight is relatively uncommon. Nationally representative data on measured weight change in US adults comes from the National Health and Nutrition Examination Survey (NHANES) I Epidemiologic Follow-up Study (NHEFS), in which body weight was measured on 2 occasions, approximately 10 years apart. Our analyses of these data show that among older adults (aged 65 years and older) with normal BMIs (18.5–24.9) in the early 1980s, only 2% had been obese (BMI ≥30) 10 years previously. Of older adults who were obese in the early 1970s, only 4% had normal BMIs 10 years later. The probability of large weight loss among obese adults is likely to be even lower today.2 Our analyses of NHANES 1999–2002 self-reported past-weight data for adults 65 years of age and older show that only 6% of non-obese older adults (BMI < 30) were obese 10 years previously and 2% of normal-weight older adults were obese 10 years previously. Hu et al. raise the same specter of reverse causation to assert that individuals with pre-existing illness should be excluded from cohorts used to derive relative risks applicable to the whole US population. However, these and other common exclusions (such as exclusion of the older elderly and of current or former smokers) not only make the remaining cohort less representative of the US population but in fact exclude the people more likely to die, thus making the deaths in the cohort even less representative of mortality in the US population. Relative risks based only on a subgroup of the US population defined by exclusions are not necessarily valid for inference to deaths outside that subgroup.