An Evidence-Based Assessment of Federal Guidelines for Overweight and Obesity as They Apply to Elderly Persons FREE

THE FIRST US Clinical Guidelines on the Identification, Evaluation, and Treatment of Overweight and Obesity in Adults,¹ released by the National Institutes of Health, National Heart, Lung, and Blood Institute, in June 1998, identified overweight and obesity as major risk factors for increased morbidity and mortality. In earlier official documents (the National Institutes of Health Consensus Conference on Health Implications of Obesity in 1985 and the 1990 Dietary Guidelines for Americans from the Department of Agriculture), overweight was defined as a body mass index (BMI; defined as weight in kilograms divided by the square of height in meters) of 27.8 or greater and 27.3 for men and women 35 years or older, respectively. Those documents recommended age-specific ranges of weight for height, with heavier weight standards indicated with increased age, and adapted the obesity and overweight terms interchangeably without clear distinction.^2- 3 In contrast, the recent national guidelines, in agreement with the 1995 Dietary Guidelines for Americans, defined overweight as a BMI of 25 to 29.9 and obesity as a BMI of 30 or more for adults 18 years or older, without recommending age-specific cut points.^1,4 According to the current definitions, the percentage of overweight and obese adults 20 years or older in the United States is estimated to be 54.9%.⁵

Although the recent recommendations consider younger and older adults as one group, they are almost exclusively based on studies that predominantly included young and middle-aged populations. Overall, there is a paucity of data relative to the association between overweight and all-cause and coronary heart disease (CHD) mortality among elderly persons and more specifically among very old individuals (≥75 years). This issue is important because the prevalence of overweight among persons who are 70 years or older is estimated to be more than 37%, and at least another 15% are estimated to be obese by the published standards.⁵

To determine the strength of evidence regarding the prognostic importance of increased weight and the appropriateness of the existing recommendations for healthy weight for older people, we performed a systematic review of the literature. We specifically focused on evidence that was relevant to people who were aged at least 65 years, with special emphasis on those who were at least 75 years old.

The sources of data for this review were published articles that examined the relation between body weight adjusted for height (BMI) and all-cause mortality, cardiovascular mortality, and CHD in individuals 65 years or older. We performed a computerized search of MEDLINE from January 1966 through October 1999 with the following search terms: body mass index/BMI/body weight/obesity, aged/elderly, and either mortality or cardiovascular disease or coronary disease/coronary artery disease/coronary heart disease. We limited our search to English-language articles.

The search resulted in 444 articles. We examined the titles and abstracts of these articles and selected those that corresponded to the inclusion criteria of our systematic review as described herein. Subsequently, we performed a detailed examination of the articles. Finally, we completed our review by examining the relevant references from the articles that had been detected by the MEDLINE search.

Our search did not identify any interventional trials regarding the topic. Observational studies were selected if they included only, or presented separate data analysis for, subjects 65 years or older; performed age adjustment; included at least 100 subjects; had at least 3 years of follow-up; had all-cause and/or cardiovascular mortality and/or CHD events as end points; restricted, stratified, or adjusted for smoking and health status at baseline; and selected nonhospitalized subjects at the time of enrollment.

We excluded studies that used weight and did not adjust for height (BMI). Studies were also excluded if they were based on specific populations of diseased individuals, eg, case-control studies of patients who presented to the hospital with acute myocardial infarction.

The end points in these studies were all-cause mortality, cardiovascular mortality, stroke mortality, cancer mortality, noncardiovascular mortality, CHD mortality, or CHD incidence. For the purpose of our systematic review, we evaluated 4 end points: all-cause mortality, cardiovascular mortality, CHD mortality, and CHD incidence.

For each article, we determined the study design, population, recruitment period, follow-up duration, number of subjects 65 years or older, age range, and sex. We recorded the methods used to assess standardized weight and the way BMI was used as a variable in the statistical models. Exclusion and inclusion criteria were evaluated for each study. We compared the methods used in the studies to control for smoking and baseline health status by exclusion, restriction, or adjustment. We also examined the factors that were controlled for in the different statistical models in addition to age, smoking, and health status.

Controversy exists with respect to the statistical adjustment for potential weight-related cardiovascular risk factors, such as high blood pressure, diabetes mellitus, and hypercholesterolemia, since those risk factors might be a consequence of excess weight and mediators of adverse outcomes rather than confounders.⁶ Consequently, we present both adjusted and unadjusted analyses for those risk factors whenever they were available. Results presented as "adjusted for potential weight-related cardiovascular risk factors" have been controlled for at least 1 of the following: serum cholesterol level; serum glucose level; systolic, diastolic, or mean arterial blood pressure; and history of diabetes, dyslipidemia, or high blood pressure.

In addition, we evaluated whether early mortality was excluded and whether adjustment was made for history of weight loss. When possible, we present the magnitude of the association between BMI and mortality and CHD and the statistical significance as relative risks (RRs) with confidence intervals (CIs) or P values for trend.

Characteristics and findings of the studies with respect to older subjects are presented in Table 1, Table 2, Table 3, and Table 4. For studies that also included participants younger than 65 years, only the information and results concerning those 65 years or older are included.

Thirteen articles met the inclusion criteria for this systematic review. The selected studies were all observational, prospective cohort studies and were population based except for one, which was based on male health professionals. Except for the Honolulu Heart Program study, which also included nursing home residents, they all included noninstitutionalized, nonhospitalized individuals. Of note, some of these investigations^7- 10 consisted of different analyses performed on data from the same observational study but with different cohorts or different recruitment periods or follow-up duration. The studies, their characteristics, and the potential weight-related covariables introduced in their statistical models are listed in Table 1 and Table 2. The studies included 1437 to 46 954 individuals, with follow-up between 3 and 23 years. Eight of these studies included both sexes, 4 included only men, and 1 included only women. Ten studies were from the United States and 3 from Europe, and except for the Honolulu Heart Program study on Japanese American men, they all included predominantly white subjects from industrialized countries. In the 3 European studies,^11- 13 the subjects' race was not reported, but it is likely that the participants from those countries (Finland, the Netherlands, and Italy) were mostly white. Only one study,¹⁰ by Cornoni-Huntley and colleagues, reported results separately for black subjects.

Weight was measured in 8 studies and self-reported in 5 (Table 2). Body mass index was used as a categorical variable in all but 1 study,¹¹ and 2 studies^14- 15 reevaluated BMI as a continuous variable. The classification of BMI varied among studies, with a range of groupings from 3 to 10 levels. Four studies^8,14- 16 controlled for history of weight loss. Four studies^7- 8,15,17 controlled for early mortality by excluding participants who died during the early years of follow-up.

Seven studies reported results on the association between BMI and all-cause mortality without adjusting for potential weight-related cardiovascular risk factors. When available, the magnitude of the association, the CIs, and the referent groups are displayed in Table 2 and Table 3.

A positive association between BMI and all-cause mortality was shown in only 2 investigations. Even though the first study, published by Harris et al¹⁸ in 1988 based on the Framingham Heart Study data, revealed a positive association between BMI and mortality for both men and women, this positive association was detected only for a BMI of 28.5 or more, and there was no significant or substantial increased risk detected for the group with a BMI between 25 and 28.5. Furthermore, Harris et al based their study on subjects who were all 65 years old at baseline; consequently, no conclusion can be made, based on this study, regarding the presence of the increased mortality risk associated with high BMI among very elderly persons (≥75 years).

The second study, by Stevens et al¹⁵ based on the American Cancer Society's Cancer Prevention data, also found that higher BMI was significantly associated with increased mortality in men and women, though exclusively for the participants aged 65 to 74 years. There was no significant association in this study between BMI and mortality for those 75 years or older. Again, for those aged 65 to 74 years, the RR was statistically significant for the BMI category of 27.0 or higher. The magnitude of the association and the CIs are not given in the article. However, the most important finding of this study was that the increased risk associated with high BMI declined with age for both men and women (P for trend <.01 and <.05, respectively). Using a BMI of 21.0 as the reference value, the BMI associated with a 20% increase in the risk of all-cause mortality was 24.7 among men aged 55 to 64 years but 28.2 and 30.5 among men aged 65 to 74 years and 75 to 84 years, respectively. Correspondingly, among women, the BMI associated with a 20% increase in the risk of all-cause mortality was 25.9 for those aged 55 to 64 years but increased to 29.9 for the 65- to 74-year-old age group.

Cornoni-Huntley et al¹⁰ found a U-shaped relation between BMI and mortality among white women aged 65 to 74 years with increased risks in the lowest (≤21.4) and highest (≥31.3) ranges of BMI. No significant association was found between BMI and mortality among white men in the same age group in this study.

All other studies showed either no association^13- 14 or negative significant associations^8,10 between BMI and all-cause mortality.

Most studies showed either a negative or a nonsignificant association between BMI and all-cause mortality. When a positive relation was found, it was for a BMI of 27 or more. Furthermore, this relation was attenuated with age and usually disappeared after the age of 75 years. Consequently, the overall trends for the relation between BMI and mortality in older adults can be represented as a U-shaped curve, with a large flat bottom and a right curve that starts to rise for BMIs of more than 31 to 32.

Despite the controversy regarding the statistical adjustment for weight-related risk factors, for completeness we included investigations that performed this adjustment. Ten studies presented results on the association between BMI and all-cause mortality after including some cardiovascular risk factors that could be weight related in their statistical models (Table 1). In general, control for weight-related cardiovascular risk factors attenuated the strength of the association between high BMI and all-cause mortality.^8,10,18

Only 2 studies demonstrated that high BMI was associated with increased all-cause mortality. Based on the Framingham Heart Study data, Harris et al¹⁸ reported a significant positive linear association between BMI and all-cause mortality. Nevertheless, this positive association was exclusively found among women who were 65 years old at entry and for a BMI of 28.7 or more. In this study, BMI was not significantly associated with all-cause mortality among men. The second study, based on the Cancer Prevention Study, found that a BMI of 28 or more was significantly related to increased mortality in the 65- to 74-year-old age group. In this study,¹⁶ the relation between BMI and mortality was found to be U-shaped, with BMIs of less than 20.5 and 28 or more associated with increased mortality, among individuals 75 years or older. Two studies^11- 12 showed a U-shaped association between BMI and mortality exclusively for people aged 65 to 74 years and no association among people 75 years or older. In one of these studies, only the extreme values of BMI (<19.0 and ≥34.0) were associated with increased mortality,¹² and the other revealed a weakly significant U-shaped relation between BMI and mortality (β coefficient, 0.0064).¹¹

Three studies^7- 8,14 showed no significant association between BMI and all-cause mortality among the elderly (≥65 years) after including some potential weight-related cardiovascular risk factors in the model. Two studies,^10,17 one published recently from the Honolulu Heart Program and the other based on the First National Health and Nutrition Examination Survey, Epidemiologic Follow-up Study, showed that only low BMI was significantly associated with increased mortality among elderly Japanese American men and among elderly men and women in both black and white populations.

With inclusion of more variables in the model, especially the potential weight-related cardiovascular risk factors, either BMI was not related to increased all-cause mortality or low BMI was associated with greater risk of mortality. When a positive association was found, it was for high values of BMI (≥28), and except in one study,¹⁶ the significant association disappeared after the age of 75 years.

Results of the association between BMI and cardiovascular mortality are less conclusive compared with those for all-cause mortality (Table 4). This association was found to be none, negative, or positive in different studies. However, when positive, the association was not strong in magnitude, started for high values of BMI (≥27-29), and did not persist after the age of 75 years.

Two investigations reported results on the association between BMI and cardiovascular mortality without adjusting for potential weight-related cardiovascular risk factors. The study by Stevens et al¹⁵ found this association to be significant for a BMI of 27 or more and exclusively in the 65- to 74-year-old age group. Again, BMI was not associated with increased mortality among persons 75 years or older. Also, the investigation¹³ among Finnish women aged 65 to 79 years did not reveal high BMI as a risk factor for cardiovascular mortality.

Three studies reported results on the association between BMI and cardiovascular mortality after adjusting for potential weight-related cardiovascular risk factors. The investigation based on the Framingham Heart Study showed a significant positive association between BMI and cardiovascular mortality among women who were 65 years old at baseline and a negative association among men in the same age group. The magnitude of the association and CIs are not reported in the article.¹⁸ Although the study by Rissanen et al¹² revealed a positive association between BMI and cardiovascular mortality among Finnish men, this association was significant for a BMI of 34 or more and exclusively among persons aged 65 to 74 years. High BMI was again not shown to be a risk factor for the very elderly (≥75 years). Finally, Seeman et al⁹ reported that BMI was not significantly related to CHD mortality.

Only 2 investigations studied the association between BMI and CHD incidence among elderly persons (Table 3).

The Health Professionals Follow-up Study¹⁹ failed to demonstrate a significant association between high BMI and CHD incidence among men 65 years or older. No association was found between weight gain in adult life and coronary risk in this group. Nevertheless, waist-to-hip ratio was a significant risk factor for CHD among men 65 years or older (RR, 2.76; 95% CI, 1.22-6.23).

A study by Seeman et al,⁹ based on Established Populations for Epidemiologic Studies of the Elderly data, showed that high BMI was significantly related to increased myocardial infarction incidence among men but not among women.

Optimum BMIs with respect to all-cause and cardiovascular mortality and CHD incidence are displayed in Table 2. In general, the nadirs for risk increased with age and, for the most part, were at high levels of BMI ranging from 27 to 30.^7,9- 15,19 Only the investigation by Cornoni-Huntley et al¹⁰ among black men and the recent study published by Calle et al¹⁶ showed an optimum BMI that was consistent with the recent guidelines (22.9-25.6 and 20.5-24.9, respectively).

Although there are not many studies on the association between BMI and mortality in elderly persons, our systematic review revealed several important findings. Most studies failed to show a significant association between high BMI and increased mortality, despite the large number of participants. In addition, while few studies found a significant U-shaped or positive linear association between BMI and mortality, only values of BMI that are higher than recent guideline cut points for overweight were associated with increased mortality.^{10- 12,15- 16,18} Even so, this association was not large, ranging from an RR of 1.15 to 1.34 for a BMI of 28.0 to 29.9 to an RR of 1.31 to 2.0 for a BMI of 31 to 35.^10,12,16 We found that only one study¹⁶ demonstrated a nadir of mortality risk among elderly subjects that was consistent with the ranges of BMI (19 to ≤25) defined as ideal weight by the new guidelines. Except for that study, this relation, when it existed, attenuated with age and seemed to disappear for persons 75 years or older.^7,11- 15,18 Finally, when the association was U-shaped, it was generally asymmetric, with a less steep curve on the right and a wide, flat bottom, which demonstrates that a broad range of increasing BMI is compatible with minimum excess mortality in elderly persons. As a consequence, the optimum BMI commonly tended to be higher for the elderly compared with young and middle-aged populations.^7,10- 15,17 In summary, the overall trends for the relation between BMI and mortality in older adults can be represented as a U-shaped curve, with a large, flat bottom and a right curve that starts to rise significantly for BMIs greater than 31 to 32.

There are even fewer studies on the association between BMI and cardiovascular mortality. However, in general, a pattern similar to the latter was found.^{9,12- 13,15,18} Results from the only 2 studies^9,19 on the association between BMI and CHD incidence are contradictory.

Authors of the recent guidelines base their recommendations on the observational and randomized clinical trial evidence. However, by necessity the expert panel used evidence from studies that mainly included young and middle-aged populations. For example, to evaluate the effects of weight loss on blood pressure, they considered 45 articles from a pool of 76 randomized clinical trials. The mean age in almost all of those studies was between 35 and 55 years, and only one study included individuals with a mean age of 61 years. Again, to evaluate the effects of weight loss on dyslipidemia, 65 articles reporting randomized clinical trials were considered for review, and 22 were accepted. The mean age in almost all of those studies was also between 35 and 55 years, and only 2 studies included individuals with a mean age of 61 years.²⁰ Even though the guidelines recommend the treatment of overweight "only when patients have 2 or more risk factors," the age itself (≥45 years for men and ≥55 years or postmenopausal for women) is considered a risk factor in this document. Therefore, based on the guidelines, a 70-year-old patient with a weight of 64 kg and a height of 160 cm and one of the following risk factors is a candidate for weight reduction treatment and interventions¹:

. . . established coronary heart disease, other atherosclerotic diseases . . . gynecological abnormalities, osteoarthritis, gallstones . . . stress incontinence . . . cigarette smoking, hypertension . . . high risk LDL-C [low-density lipoprotein cholesterol] . . . low HDL-C [high-density lipoprotein cholesterol], impaired fasting glucose level . . . family history of premature coronary heart disease . . . physical inactivity, and high serum triglyceride levels.

The methodologic challenges of isolating the impact of weight on prognosis are considerable. The issues in these studies include lack of control for smoking status; preponderant effects of poor baseline health status and underlying diseases of underweight participants, hampering the negative effects of higher BMI on mortality (hence, impeding high BMI to emerge as a risk factor for excess mortality); inappropriate control for some potential weight-related conditions (high blood pressure, diabetes, and dyslipidemia) that would weaken the association between high BMI and mortality; and lack of power to detect a significant association even if one existed. We considered these methodologic problems in our selection of studies. All the studies that we reviewed were controlled for smoking status. Although we recognize that underweight could be a consequence rather than a cause of poor health that might result in excess mortality, all selected studies in our systematic review were controlled in some way for baseline health status. All the study populations, except for that in the study by Kalmijn et al,¹⁷ consisted of noninstitutionalized, relatively healthy elderly individuals. In addition, the inclusion criteria in most of those studies were directed to minimize the effects of preexisting diseases on the outcomes. Studies that excluded participants with important weight loss before the study baseline^8,14- 15 and/or early mortality^7- 8,15 found a decline in association between low BMI and mortality but did not show increased risk of mortality associated with overweight. This was also true for investigations^10,13 that performed separate analyses for early and later duration of follow-up. Therefore, the absence of prognostic significance of increased weight observed in those investigations is unlikely to be a consequence of the excess mortality in low-weight participants.

Body mass index was used as a categorical variable in most of these studies, and the RR was calculated compared with the reference group with the middle or healthy range of BMI and not with the lowest BMI category or nonobese persons in general.^{7- 8,10,12- 13,15,18} We also looked at the issues concerning the importance of unadjusted association between BMI and mortality rather than the independent association and presented both results from unadjusted and adjusted analyses for weight-related cardiovascular risk factors whenever they were available. In general, control for these risk factors attenuated the RR of mortality associated with high BMI.^8,10,18 Also, inclusion of chronic conditions and history of weight loss in the model weakened the strength of association between low BMI and mortality.^8,14- 15 Finally, all the studies that we reviewed were observational studies that included a large number of participants.

How is it possible that mild-to-moderate excess weight is not an important risk factor for cardiovascular disease and all-cause mortality in elderly persons in contrast with young and middle-aged individuals? It is conceivable that overweight individuals who survive to old age have characteristics that protect them from adverse effects of overweight. Also, underweight-related risk factors might prevail over the risks related to overweight in old age. It is possible that the significance and prognostic effects of traditional cardiovascular risk factors may change with increased age. For example, several studies found that elevated total serum cholesterol and LDL-C levels or low HDL-C levels are not associated with increased overall and cardiovascular mortality or first myocardial infarction in elderly patients^21- 23 and more specifically in very elderly patients.²⁴ Data from the Framingham Heart Study revealed that the positive relation between high cholesterol level and all-cause and cardiovascular mortality declined with age to become negative after the age of 80 years.^25- 26 The interaction of age on the relation between weight and health outcomes is not a recent finding. Various studies that included subjects from different age groups reported that the association between BMI and mortality and CHD (all-cause and cardiovascular) declines with age.^27- 39 A recent study⁴⁰ among obese patients in Germany (age, 18-74 years; BMI, 25 to ≥40) with a median follow-up of 14.8 years also showed that the excess mortality associated with obesity declined with age and that the lowest mortality risk was observed in patients aged 50 to 74 years with BMIs of 25 to less than 32.

Interestingly, we found that low BMI was more consistently associated with greater mortality risk compared with high BMI in elderly persons. Different factors appear to prevail for longevity in older people. The protective effects of overweight, such as nutritional reserve, may prevail over its negative effects in this population. Malnutrition, osteoporosis, and traumatic events secondary to fall (specifically hip fracture and mortality allied with those conditions) could be an explanation for the increased mortality associated with weight loss and underweight.^41- 43

The absolute gain from modifying risk factors and the risk-benefit balance with respect to interventions are complicated to evaluate in elderly persons. In contrast with younger persons, elderly persons are usually faced with multiple health hazards and shorter life expectancy. Even if the mechanisms of disease are the same for young and older people, the multiplicity of risk factors accompanied by increased morbidity modify the prognostic significance of each individual factor; hence, the concept of competing risks becomes more relevant with age.⁴⁴ Consequently, a single cause-specific mortality affects the survival less as the patient becomes older. Also, the absolute benefit of reducing a specific risk depends on the magnitude of the baseline RR attributed to this hazard. That is, since an elderly person has several risk factors for so many adverse outcomes, the relative gain in survival with changes in 1 risk factor is often small unless this risk factor has a large impact on all-cause mortality.

We performed a comprehensive search of all the studies on the association between BMI and mortality and CHD. We also conducted an extensive analysis of the literature with respect to this issue, including other review articles. To our knowledge, almost all the reviews published to date are narrative. Our goal was to achieve a systematic review with predesigned methods and predefined selection criteria. We performed a systematic review and not a meta-analysis because of the heterogeneity of the observational studies. We examined the articles objectively and completed a general synthesis of the existing evidence with regard to the issue. We also considered the methodologic problems concerning the impact of smoking status, underlying disease, and baseline health status in our selection of studies.

We were restrained by the limitations of the quality of the observational studies, which do not provide definitive evidence about causality. Data in most of these studies were collected based on self-coded interviews, and weight, height, and comorbidity are usually self-reported. The end points in these investigations were limited to mortality (all-cause and cardiovascular) and CHD incidence. However, mortality is not the sole indicator of good health. Quality of life and functional and psychosocial status are also important outcomes that should be taken into consideration. In addition, little is known about the influence of sex, race, and weight history on the association between BMI and mortality.

Evidence does not support mild-to-moderate overweight, defined by the new guidelines, as a risk factor for all-cause and cardiovascular mortality among elderly persons. We agree that marked overweight (obesity) might be a risk factor for this population. However, adverse effects of energy (calorie) restriction in elderly persons and the potential harms of diet-induced weight reduction, as well as the paucity of information about effectiveness of weight loss programs, would not support weight reduction interventions among mildly to moderately overweight elderly individuals. Further studies that focus on elderly and very elderly individuals are required, and future guidelines should consider age-specific recommendations that reflect existent evidence when establishing standards for ideal weight.

Accepted for publication September 21, 2000.

We thank Maria Johnson for her outstanding editorial assistance.

Corresponding author and reprints: Harlan M. Krumholz, MD, Yale University School of Medicine, 333 Cedar St, PO Box 208025, New Haven, CT 06520-8025 (e-mail: harlan.krumholz@yale.edu).