Center for Research on Health Care, Division of General Internal Medicine, 230 McKee Place, Suite 600, Pittsburgh, Pennsylvania 15213. E-mail: email@example.com
Objective: Although obesity is increasing in older U.S. adults, treatment is controversial in this age group. We sought to examine evidence concerning obesity's health-related risks, diagnostic methods, and treatment outcomes in older individuals.
Research Methods and Procedures: We searched MEDLINE and Cochrane Library databases, consulted with experts, and examined bibliographies for English language studies discussing obesity in older adults (mean age ≥ 60), published between January 1980 and November 2005. Inclusion criteria were met by 32 longitudinal analyses, seven diagnostic studies, and 17 randomized controlled trial articles. At least two authors independently reviewed and abstracted study design, population, results, and quality information.
Results: Correlations between body fat and three anthropometric measures (BMI, waist circumference, waist-to-hip ratio) decrease with age but remain clinically significant. Obesity contributes to risk for several cardiovascular endpoints, some cancers, and impaired mobility but protects against hip fracture. The association between obesity and mortality declines as age increases. Intensive counseling strategies incorporating behavioral, dietary, and exercise components promote a weight loss of 3 to 4 kg over 1 to 3.3 years. The loss is linked with improved glucose tolerance, improved physical functioning, reduced incidence of diabetes and a combined hypertension and cardiovascular endpoint, and reduced bone density.
Discussion: In older adults, obesity can be diagnosed with standard clinical measures. Intensive counseling can promote modest sustained weight loss, but data are insufficient to evaluate surgical or pharmacological options. Obesity treatment is most likely to benefit individuals with high cardiovascular risk. Limited data suggest possible functional improvement. Treatment should incorporate measures to avoid bone loss.
In 1999 to 2000, 33% of men and 39% of women 65 to 74 years old were obese (1). Although prevalence is rising (1), and obesity is associated with major causes of death and diverse morbidity (2,3, 4,5), its treatment in elders has generated controversy. For example, some cite concerns that food restraint may lead to malnutrition or adverse muscle and bone effects (e.g., sarcopenia) and suggest that moderate obesity provides a metabolic reserve against disease. Others note that excess weight is linked with common, potentially preventable disorders in older adults and assert that health promotion remains important in old age (6, 7, 8).
Guidelines for obesity management often reflect data from young and middle-aged adults (2)(9)(10)(11) or do not use the preferred (12)(13)(14) systematic approach to evidence synthesis (15). Rates of death from obesity-related disorders and functional limitations increase with age (16)(17), but the net effect of aging on weight-related health risk is unclear. This risk may change as fat distribution shifts with age (18)(19)(20)(21) because both the amount (2) and location (22)(23)(24)(25)(26)(27)(28)(29)(30)(31)(32)(33) of fat affect risk. Shifts in fat distribution may also influence the accuracy of diagnostic tests for obesity. Likewise, age-related physiological changes may influence whether obesity interventions lead to weight loss or improve health.
To better understand obesity-related health risk and the potential for interventions to alter such risk, we systematically reviewed the medical literature on obesity in older adults. Because physical activity interventions were not considered policy-relevant but have important clinical implications for obesity diagnosis and treatment, here we update and supplement our earlier review (34) with evidence regarding physical activity.
Research Methods and Procedures
Based on an analytic framework for obesity screening in the general adult population (35), we developed four key questions regarding obesity in older adults: Is obesity associated with long-term health risks? Can obesity be diagnosed with anthropometric measures? Can intervention result in weight loss? Can weight loss result in improved health outcomes? We searched for evidence supporting these questions.
We used standard criteria to rate studies in terms of strength of evidence and internal validity (36)(37). To capture all data relevant to the aging population, we included samples with average age ≥ 60 years. Because the amount of weight lost through interventions peaks at 6 months and losses may be transient (2)(10), we excluded randomized controlled trials (RCTs)1 with follow-up of less than 1 year.
To determine long-term health implications of excess body weight, we examined observational studies with at least 10 years of follow-up. Our internal validity evaluation for these studies included methodological issues that influence estimates of weight-related health risk. We preferred studies in which results were adjusted for smoking or analyses performed separately for never-smokers and ever-smokers. We likewise preferred analyses that did not adjust for obesity-related risk factors such as diabetes or hypertension. Whenever possible, we focused on estimates unadjusted for diet and exercise because these behaviors mediate obesity development, and we were interested in the net effect of obesity on health. We considered a comparison of risk between adults with normal body weight and those with elevated body weight to be the best method to estimate health effects of excess weight, particularly in cases in which health risk showed a U-shaped pattern with increasing body size. This focus not only emphasized our primary question but also helped minimize the possible role of reverse causation because low body weight may reflect underlying disease processes with associated mortality risk, including subclinical disease [of particular significance because morbidity often presents with weight loss in the elderly (38)(39)]. We also attempted to minimize the effects of this possible bias by focusing on studies that accounted for participants’ baseline health status and abstracting the effect of excluding early outcomes from analyses (40); however, we recognize that the latter technique's overall effect may be minimal (41).
To examine potential office-based diagnostic tests for generalized obesity [BMI (kilograms per meter squared)] or central obesity [waist circumference (WC; centimeters) and waist-to-hip ratio (WHR)], we identified studies that compared at least one of these measures with an alternative estimate of body fat (BF).
To evaluate the effects of intentional weight loss, we used data only from RCTs (36) because we wanted to avoid the potential confounding by unintentional weight loss [a common occurrence with disease or psychological distress in older adults (42)] or difficulty in measuring volition that may occur in observational studies (19)(43)(44). We were primarily interested in whether obesity interventions reduce mortality and morbidity and improve quality of life and function. However, most RCTs have insufficient follow-up duration to evaluate long-term outcomes (e.g., diabetes, cancer, and cardiovascular events). Therefore, we also studied intermediate cardiovascular outcomes, including glucose control, blood pressure, and lipid levels (45)(46)(47).
We searched MEDLINE and Cochrane Library databases, consulted with experts, and examined bibliographies of English language reviews and articles published between January 1, 1980, and November 18, 2005. To find additional articles, we examined evidence tables from earlier systematic reviews (2)(10)(11)(48), consulting studies from which the tables were derived when necessary. At least two investigators independently examined each publication and abstracted data from publications meeting inclusion criteria.
We reviewed the abstract or text of over 3000 publications. Of the 56 that met inclusion criteria, 32 examined weight-related health risk, seven examined diagnostic tools, and 17 examined interventions.
Obesity-Related Health Risks
The 32 longitudinal studies of health risk (see online supplement, available at the Obesity web site, www.obesityresearch.org) used two methods to assess weight-health relationships: comparison of the risk in each body size category with the risk in a referent body-size range and/or assessment for trends in risk across the baseline range of body size. For studies using the first method, we report results stratified according to referent categories. Most studies did not report absolute risk; when reported, it typically showed the same pattern as relative risk (RR). Some studies adjusted for direct mediators of obesity (diet, exercise, or both) (22)(49)(50)(51)(52)(53)(54), so they may underestimate obesity's net health effect. Most studies did not exclude early outcomes.
Most studies examining incident cardiovascular-related morbidity (risk factors or endpoints) found that risk significantly increased with BMI (Figure 1) (22)(44)(55)(56)(57)(58). Studies examining WC or WHR and cardiovascular morbidity (22)(55)(56) yielded similar results.
In a cohort of U.S. women, cancer of any type was 20% more common in the largest BMI vs. smallest BMI participants (22). An elevated BMI was associated with increased risk of leukemia (54) and breast or uterine cancer (22) among women. It was significantly linked with colon cancer among men (53) and showed either significant (22) or borderline significant (53) associations with colon cancer among women. Incident lung cancer was less common in obese individuals (22)(51). Incident prostate cancer (50)(59), non-Hodgkin's lymphoma (60), and ovarian cancer (22)(61) were not linked to body size. Studies examining WC or WHR and cancer incidence (22)(51)(60) yielded similar results.
Women with elevated BMI were twice as likely as leaner women to develop impaired mobility (62). Men with an elevated BMI were more likely than leaner men to initiate long-term medication (58). An elevated BMI was protective against hip fracture (22) and was not associated with incident Parkinson's disease (63).
The relationship between BMI and all-cause mortality generally followed a U-shaped pattern, with highest risk among underweight and obese individuals. Accordingly, we focused on the four studies in which the referent group consisted only of individuals with normal BMI (49)(58)(64)(65). Doing so emphasizes the comparison of interest and minimizes the effect of possible reverse causation in the BMI-mortality relationship.
Among these four studies, three examined never-smokers (49)(58)(64) and one examined non-smokers (65), and only one of these studies excluded early deaths (65). Two of the analyses accounted for both smoking and baseline comorbidity status and did not overadjust for weight-related cardiovascular risk factors in RR estimates; they were also based on large samples and represent the strongest data for establishing a relationship between excess body weight and mortality (49)(64).
Both of these studies (49)(64) showed increased mortality risk among the largest individuals, an association that lessened in strength with age. Among the oldest individuals (≥75 years), one study showed mortality risk persisting for men and women (RR, 1.4 to 1.5) (64), and one showed it persisting only for men (49) (although in additional analyses, women's obesity-related mortality risk extended through age 79) (66). An age-related reduction in risk was also noted in a review incorporating shorter term studies (67). Of the normal-BMI referent studies with more methodological concerns, one showed a significant association between increasing body size and mortality (65), and one did not (58).
In the four studies that included overweight individuals in the referent group (potentially assessing an insufficient BMI gradient to detect altered health risk) (52)(65)(68)(69), most estimates of mortality risk were not significantly increased. The exceptions were increases reported in a Swedish sample (52) and a white female subset of a U.S. sample (68). In one sample in which the largest BMI category's mortality did not differ significantly from the referent's mortality, a significant trend for increasing risk was noted across the BMI range (69).
In the five studies that included underweight individuals in the referent group, we could not assess the change in risk between individuals with normal BMI and those with elevated BMI, but the mortality risk among the largest individuals was either similar to that among the referents (22)(25)(70)(71) or was lower (71)(72). Although these studies all adjusted for smoking status and tried to exclude major comorbidities, their referent groups are likely to disproportionately include individuals with subclinical disease. In addition, among these studies, two controlled for a weight-related cardiovascular risk factor in assessing BMI-mortality relationships (22)(71). In the few studies that examined long-term all-cause mortality associated with central obesity (22)(25) (see online supplement, available at the Obesity web site, www.obesityresearch.org), WC and WHR showed higher point estimates of RR than did BMI and mixed significance.
Analyses of cause-specific mortality were limited by sample size. There was no consistent relationship between cardiovascular mortality and BMI (22)(25)(40)(49)(65)(70)(71)(72)(73)(74)(75). The single normal-BMI referent study found a significant positive association, which became non-significant after age 74 in women [overlapping graphic confidence intervals (CIs) could not be assessed in men] (49). WC or WHR showed stronger positive associations with cardiovascular mortality than did BMI in two studies (22)(25).
We found little evidence for obesity-related cancer mortality (22)(25)(65)(72)(73)(76). One study showed increased all-cancer mortality among adults with elevated WHR (22), and another showed reduced lung cancer mortality among heavier participants (76). Elevated BMI showed mixed (25)(65) associations with respiratory disease mortality and was protective against hip fracture mortality (an effect that appeared in the normal BMI range) (77).
Diagnosis of Obesity by Anthropometric Measures
We identified five studies that compared anthropometric measures with alternative non-invasive BF estimates (78)(79)(80)(81)(82) (see online supplement). When BF was measured with both DXA and bioelectrical impedance, we present the DXA information because it is preferred in elderly individuals (83).
Although the correlation between BMI and BF percentage declined with age (80), most studies showed a reasonably high correlation (0.73 to 0.93), even in quite elderly samples (78)(79)(81). A single study showed a quite low correlation in men (0.37) and women (0.51) >65 years old (80).
The correlation between WC and BF percentage was fairly high (0.64 to 0.78) (81), whereas that between WHR and BF percentage was low (79)(81). In one study, elevated WC had high specificity (97% to 100%) but low sensitivity (33% to 64%) for detecting generalized obesity (elevated BMI) or central obesity (elevated WHR) (84).
Two studies explored other anthropometric differences. One determined gender- and race/ethnicity-specific WC cut-off points corresponding to a BMI of 30 and found that the derived WC values did not differ appreciably by race/ethnicity (85). The other found that among Asians, a relatively low BMI corresponded with high BF (>40% in women and >30% in men) (82).
Efficacy of Weight Loss Intervention
We identified 11 different counseling-based RCTs meeting eligibility criteria, with 12 to 48 months of follow-up (Table 1). Several reported outcomes in multiple articles. Most involved U.S. samples (86)(87)(88)(89)(90)(91) with participants who were generally healthy and in their 60s. Internal validity was fair (88)(90)(91)(92)(93)(94)(95)(96) to good (86)(87)(89)(97)(98)(99)(100)(101)(102).
Table 1. Efficacy of counseling-based interventions in older adults*
Mean weight change in intervention group vs. control group†
DPP, Diabetes Prevention Program; D, dietary intervention; E, exercise intervention; B, behavioral intervention; CI, confidence interval; TONE, Trial of Non-pharmacologic Intervention in the Elderly; HR, hazard ratio; NS, not statistically significant (p ≥ 0.05); ADAPT, Arthritis, Diet, and Activity Promotion Trial; SF-36, Short Form 36; WOMAC, Western Ontario and McMaster Universities Osteoarthritis Index; HbA1c, hemoglobin A1c; NR, not reported; PATH, Physical Activity for Total Health; Δ, change.
Data reflect randomized controlled trials with fair-to-good internal validity; validity concerns included high attrition, potential confounding, and a lack of adverse effect reporting. Multiple articles reflecting data from a single randomized controlled trial are labeled with the study name in addition to the primary author.
If a study's intensity crossed multiple categories, [e.g., low-to-moderate (94) or moderate-to-high ((90)], it was included under the lower category in the table.
Persons not requiring reinstitution of antihypertensives: 23% of the weight loss + low-sodium group, 17% of the weight loss group, 15% of the low-sodium group, and 7% of the usual care group (p = 0.012). Intervention had an NS effect on incident coronary heart disease.
Reduction in the sum of serum blood glucose values before and after oral glucose tolerance testing (−3.6 mM) in the intervention vs. control groups (p = 0.008)
We categorized interventions according to counseling frequency, with moderate-intensity interventions occurring monthly during the first 3 months, low-intensity interventions occurring less often, and high-intensity interventions occurring more often. Intensity was high in seven trials (87)(89)(91)(93)(95)(96)(98), moderate to high in one trial (90), moderate in one trial (99), moderate to low in one trial (94), and unreported in one trial (92). We also categorized interventions according to components (diet, exercise, and behavioral techniques).
Of the high-intensity interventions with diet and exercise components, two demonstrated significant weight loss (3 to 4 kg more than control at 18 to 30 months) (87)(101), one demonstrated weight loss of borderline significance (2 kg more than control) (91), and one found that adding resistance training to a weight loss intervention did not promote additional weight loss (95). A high-intensity exercise intervention without diet did not promote significant weight loss (93)(98). One study found no weight loss difference between high- and low-protein diets (96).
Of the four RCTs without high-intensity interventions (90)(92)(94)(99), only one showed significant weight loss (−1.4 kg) (99). A trial that compared usual care, an intensive intervention, and a less intensive intervention that was adapted to Medicare's reimbursement guidelines found no significant weight loss in the less intensive arm (91). Two of the trials showing no significant loss (90)(92) are difficult to interpret because of potential confounding by smoking cessation.
Of the six RCTs including a behavioral component that reported weight change in older individuals (87)(90)(91)(94)(99)(101), four showed a weight loss that was significant (87)(99)(101) or of borderline significance (91). Of the four RCTs without a clear behavioral component (92)(93)(95)(96), none showed a significant treatment effect on weight loss. The effective dietary interventions typically focused on reducing caloric intake (103), often with reductions in intake of saturated fats and cholesterol (89)(90)(92)(94). Exercise programs typically promoted moderate physical activity.
No RCTs evaluated interventions specifically for maintenance of weight loss. However, in the Trial of Non-pharmacologic Interventions in the Elderly (TONE) sample, at 4 years of follow-up (∼2 years after intervention was discontinued), weight change among weight loss participants was no longer significantly different from that among controls (97).
We identified no RCTs evaluating surgical or pharmacological obesity treatment in older adults. In younger adults with severe obesity (BMI > 40 or BMI ≥ 35 with health complications) that was unresponsive to conservative management, bariatric surgery has led to marked weight loss (2)(10)(11)(35)(48)(104) and has been associated with relatively low rates of adverse surgical outcomes. However, these results may not generalize to older adults, who typically have a higher operative risk (105)(106)(107).
Health Consequences of Lifestyle Intervention
In several intervention RCTs, investigators examined health outcomes in addition to weight loss (Table 1). One study found that an average weight loss of 3.0 kg corresponded 2.5 years later with a 30% reduction in a combined cardiovascular endpoint (cardiovascular events, poorly controlled blood pressure, and the need to re-initiate anti-hypertensive medications) in black and white participants (86)(87). One intensive lifestyle intervention reduced diabetes incidence by 58% in the overall sample (mean age, 51 years) and by 71% in individuals ≥ 60 years (89). A high-intensity intervention with borderline significant weight loss did not alter hemoglobin A1c (HbA1c) (91). A moderately intense exercise program improved insulin levels and insulin resistance but not fasting glucose or triglyceride levels (102), and a low-intensity intervention that did not change weight reduced serum glucose (92). A high-intensity high-protein (vs. low-protein) intervention improved blood pressure but not markers of glycemic control or cholesterol (96). One exercise intervention without an effect on weight did improve oxygen uptake (93), whereas a low-to-moderate intensity program altered neither weight nor serum cholesterol (94). After 4 years of follow-up, although weight loss no longer persisted, 10% to 16% more of the TONE weight loss participants vs. controls continued to require no anti-hypertensive medication (97).
In an RCT involving older adults with arthritis, lifestyle intervention led to a small but clinically relevant increase in physical function (98)(108), less pain, and improved mobility and self-efficacy (100)(101). Likewise, adding resistance training to weight loss intervention improved muscle strength (95).
One RCT found that weight loss was accompanied by bone mineral density loss (88) at a rate that is unlikely to be clinically meaningful over the range of weight typically lost with counseling interventions but may be relevant for very old individuals given the age-related increase in hip fracture risk (109). The use of high- vs. low-protein diets did not influence bone density (96).
Summary of Findings
Obese older adults are at increased risk for incident cardiovascular risk factors and events, certain cancers, and impaired mobility but are somewhat protected from hip fracture and lung cancer. The link between obesity and all-cause mortality diminishes with increasing age and is greatly reduced or absent by the time individuals reach their mid-70s to early 80s.
In older adults, obesity can be diagnosed easily and inexpensively using anthropometric measures. Although the correlation between BMI and fat mass is lower in older adults than in the general population, the difference is typically small. BMI may have the greatest clinical utility because it is linked with the widest range of health states, and WC and WHR may be useful adjuncts for assessing cardiovascular risk. The utility of diagnostic tests may differ by race/ethnicity.
In older adults, intensive counseling-based interventions incorporating diet and exercise can lead to modest (3 to 4 kg) sustained weight loss. This loss is sufficient to show a clinical improvement in glucose control and a reduction in the combined incidence of blood pressure abnormalities and cardiovascular events. Intervention results in a similar amount of weight loss in younger adults, but this amount appears to be more beneficial in preventing or delaying the onset of diabetes in older individuals. Modest weight loss may improve overall physical functioning in older adults with lower extremity arthritis. However, behavioral intervention may promote reduced bone density. Studies, to date, have not measured the effect of weight loss intervention on mortality or cancer outcomes.
Data regarding counseling-based interventions typically reflect the use of low-calorie, relatively low-fat diets. Most successful counseling-based interventions included moderate physical activity, which is important because it can reduce the risk of bone density loss (88), minimize the risk of sarcopenia (110), and mitigate other adverse health effects of obesity (111)(112). As with younger adults (35), the use of interventions that have multiple components (e.g., diet, exercise, and behavioral counseling) tend to be the most successful. Studies generally have not assessed the harms of counseling-based interventions.
Data are insufficient to evaluate the safety or efficacy of pharmaceutical or surgical approaches to weight loss in older adults. Because the prevalence of chronic illness increases with age, and because both age and comorbidity are linked with perioperative risk (105)(106)(107), rates of adverse surgical outcomes found in younger adults may not be generalizable to older adults. Limited observational data suggest that bariatric surgery can be safe in the short term in older adults (113), but a systematic evaluation of the data generalization is beyond the scope of this review.
This review is limited by the relative lack of data specific to older adults, especially very old adults. Longitudinal study limitations included high attrition rates, inappropriate adjustment for potential confounders, suboptimal choice of referent group, and inconsistent adjustment for early mortality. Diagnostic test studies often lacked testing parameters. In studies of counseling-based interventions, negative intervention-related issues were minimally reported, attrition was often high, and samples lacked diversity in ethnicity, health status, or degree of excess weight. No RCTs focused on weight loss maintenance. Surgical and pharmacological weight loss approaches lacked evidence of efficacy in older adults.
Relatively healthy older adults who are at increased risk for cardiovascular disorders or arthritis-related functional impairment are likely to benefit from diagnosis of obesity and initiation of intensive lifestyle interventions. Intensive interventions that include diet, physical activity, and behavioral components are most likely to promote health. Dietary components should reflect evidence-based approaches for weight loss while emphasizing the importance of nutrition, because malnutrition is common in the elderly and sometimes difficult to detect (114)(115). Physical activity should be tailored to accommodate chronic disease, sensory deficits, or functional limitations. In older adults at high risk for osteoporosis, any consideration of intentional weight loss must carefully balance the potential benefits and harms and should incorporate physical activity to minimize bone loss. Although weight management may prevent, delay, or lessen the need for pharmacological treatment in older adults with cardiovascular risk factors, it is unlikely to eliminate pharmacotherapy. Innovative approaches may be needed to reduce current barriers (e.g., stigma, lack of evidence-based programs, high costs of existing programs, and lack of health insurance coverage for programs) to counseling-based interventions in older adults.
This work was supported by The Agency for Healthcare Research and Quality and Centers for Medicare and Medicaid Services and by the University of Pittsburgh, Division of General Internal Medicine. The federal funders were involved in the design of the study, assisted in the collection of some of the articles identified for review, organized an external peer review of the original technology assessment, and reviewed preliminary versions of that manuscript. They were not involved in the conduct of the study, abstraction, management, analysis or interpretation of the data, or preparation or approval of the manuscript. The authors of this article are responsible for its contents, including any clinical or treatment recommendations. No statement in this article should be construed as an official position of the Agency for Healthcare Research and Quality or of the U.S. Department of Health and Human Services.
Nonstandard abbreviations: RCT, randomized controlled trial; WC, waist circumference; WHR, waist-to-hip ratio; BF, body fat; RR, relative risk; CI, confidence interval; TONE, Trial of Non-pharmacologic Interventions in the Elderly; HbA1c, hemoglobin A1c.
The costs of publication of this article were defrayed, in part, by the payment of page charges. This article must, therefore, be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.