| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
ORIGINAL RESEARCH COMMUNICATION |
1 From the Centre for Ageing and Public Health (AEF, GMP, and EB) and the Nutrition and Public Health Interventions Research Unit (RU), London School of Hygiene and Tropical Medicine, London, United Kingdom, and the Section of Care of the Elderly, Faculty of Medicine, Imperial College, Hammersmith Campus, London, United Kingdom (CJB)
2 Supported by the UK Medical Research Council, Department of Health for England and Wales and the Scottish Office (to the MRC Trial of Assessment and Management of Older People) and by the Health Foundation, London. 3 Reprints not available. Address correspondence to AE Fletcher, Centre for Ageing & Public Health, London School of Hygiene and Tropical Medicine, Keppel Street, London WC1E 7HT, United Kingdom. E-mail: astrid.fletcher{at}lshtm.ac.uk.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Objective: We investigated the association of BMI (in kg/m2), waist circumference, and waist-hip ratio (WHR) with mortality and cause-specific mortality.
Design: Subjects aged 
75 y (n = 14 833) from 53 family practices in the United Kingdom underwent a health assessment that included measurement of BMI and waist and hip circumferences; they also were followed up for mortality.
Results: During a median follow-up of 5.9 y, 6649 subjects died (46% of circulatory causes). In nonsmoking men and women (90% of the cohort), compared with the lowest quintile of BMI (<23 in men and <22.3 in women), adjusted hazard ratios (HRs) for mortality were <1 for all other quintiles of BMI (P for trend = 0.0003 and 0.0001 in men and women, respectively). Increasing WHR was associated with increasing HRs in men and women (P for trend = 0.008 and 0.0002, respectively). BMI was not associated with circulatory mortality in men (P for trend = 0.667) and was negatively associated in women (P for trend = 0.004). WHR was positively related to circulatory mortality in both men and women (P for trend = 0.001 and 0.005, respectively). Waist circumference was not associated with all-cause or circulatory mortality.
Conclusions: Current guidelines for BMI-based risk categories overestimate risks due to excess weight in persons aged 75 y. Increased mortality risk is more clearly indicated for relative abdominal obesity as measured by high WHR.
Key Words: Older persons • obesity • mortality • body mass index • waist-hip ratio
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
SUBJECTS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Oral informed consent was obtained from all participants. Local research ethics committee approvals were obtained for each participating practice.
The assessment covered a wide range of physical, social, and psychological problems. Anthropometric measures included height, weight, and waist and hip circumferences. Participants were asked to remove all clothes except their undergarments and to take off their shoes. Height was measured to the nearest 0.1 cm by using a portable Microtoise stadiometer (CMS Weighing Equipment Ltd, London, United Kingdom) that was fixed to a wall. Weight was measured by using Soehnle scales (Soehnle-Waagen GmbH & Co, Murrhardt, Germany) and recorded to the nearest 0.1 kg. Participants were asked to roll down their undergarments to permit waist and hip measurements. Waist was measured to the nearest 0.1 cm midway between the iliac crest and the lower rib margin by using an insertion tape. Hip was measured as the widest circumference over the buttocks and below the iliac crest. Height and weight were measured only once. Duplicate measurements were made in all other cases, and the average was used for analysis unless there was > 3% difference between duplicates, in which case the measurement was not used for analysis.
The UK Office for National Statistics provided the date and the coded cause of death according to the International Classification of Diseases, 9th revision (ICD-9), for deaths reported up to September 2002—86% of the deaths in this analysis—or 10th revision (ICD-10) after that date. Analyses are based on deaths up to 31 July 2003 for all-cause mortality; underlying causes were coded as circulatory (codes 390–459 in ICD9 and I1-I99 in ICD10), cancer (codes 140–209 in ICD9 and C1-C99 in ICD10), and respiratory (codes 416-519 in ICD9 and J1-J99 in ICD10) causes.
Statistical analysis
Data management and analysis were conducted by using STATA software [version 8.0; Stata Corp, College Station, TX (24)]. BMI and WHR indexes were calculated. The outer 0.27% (> 3 SDs) of anthropometric values and indexes were excluded. Anthropometric data were categorized into quintiles by sex-specific quintiles and related to mortality risk by using Cox regression models. Three-way interactions were tested between each anthropometric variable (BMI, WHR, and WC) with sex and current smoking status for all-cause and circulatory mortality.
All models were adjusted for the linear effects of height and age, which was considered the basic model (model 1 in the tables). In addition, potential confounders were identified a priori from the detailed health assessment (model 2 in the tables); these variables included psychosocial factors (serious illness in or bereavement for loved one in the previous year, depression (score of 6 on the Geriatric Depression Scale; 25), cognitive impairment (score of 
17 on the Mini-Mental State Examination; 26; or of < 12 when the Language, Writing and Drawing performance items could not be completed), socioeconomic factors (housing tenure categorized as owner-occupier, rental, retirement home, or assisted living, and Carstairs local area deprivation score categorized by UK quintiles of distribution; 27), former smoking (in models excluding current smokers), recent alcohol use (total units drunk in past wk), and unexplained recent weight loss (> 3.2 kg in the previous 6 mo). Analyses were conducted with additional adjustment (model 3 in the tables) with a set of covariates conjectured a priori to be at least potentially or partially associated with body composition and therefore to be possible pathway variables. These covariates included self-reported history of cancer, heart attack, stroke, diabetes, or respiratory disease; the number of falls at home in the previous 6 mo; concurrent angina symptoms and respiratory symptoms, including persistent cough, wheeze, and shortness of breath while walking; sitting systolic blood pressure (average of 2 readings); physical activity (not at all, a little, fairly much, or very much); and the number of activities of daily living (ADLs)—ie, washing self, dressing self, cutting toenails, cooking, shopping, doing light housework, walking 50 yards (45.7 m), and going up and down stairs and steps—that the subject was unable to do. Persons with cancer at baseline were excluded from analyses of cancer mortality. Additional analyses explored the effects of excluding the first 1 and 2 y of follow-up and restricting the analyses across the full follow-up period to a subset of participants defined as healthy (ie, no history at assessment of cancer, heart attack, stroke, diabetes, Parkinson's disease, hip fracture, angina, weight loss, pneumonia or leg ulcer; not cognitively impaired; able to carry out 1 ADL; < 2 falls in previous 6 mo; and self-reported physical activity reported as better than "not at all"). Robust SEs were used to take account of the study design of 53 practices (28) by using the cluster (ie, practice) option for Cox regression and the Survey functions in STATA software for other regression models or crosstabulations. Robust SEs were also used to calculate 95% CIs, and design-adjusted Wald tests of significance were reported for regression models.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
75 y were eligible for a detailed assessment. Of these, 5811 (74%) men and 9349 (68%) women participated in the study, and 5715 men and 9118 women underwent 1 anthropometric measurement. One-third of assessments were done in the participant's own home, and the rest were done in the family practitioner's office (65%) or in a retirement or assisted living residence (2%). Participants for whom there was 1 anthropometric measure had a lower risk of subsequent mortality [age- and sex- adjusted hazard ratio (HR) = 0.72; 95% CI: 0.68, 0.76] than did nonparticipants or participants with no anthropometric data. BMI was unavailable in 1049 subjects because of missing measurements of height (n = 594), weight (n = 152) or both (n = 303); 910 subjects did not have WHR data because of missing measurements for WC (n = 133), hip circumference (n = 237), or both (n = 540). Only a small number were excluded because of a difference of > 3% in repeat measures: 87 subjects for WC and 78 for hip circumference. In men, there were significant negative trends across the quintiles of BMI for an association with age, height, proportion of current smokers, unexplained weight loss, depression, and cognitive impairment, whereas diabetes and systolic blood pressure were positively associated (Table 1
). WHR showed similar trends but not for depression and cognitive impairment. WHR in men also was positively associated with weekly alcohol consumption, history of cardiovascular (heart attack or stroke) or respiratory disease, and low physical activity. In women, some differences in the opposite direction were observed for WHR compared with BMI with respect to associations with age, alcohol consumption, and cognitive impairment (negative with BMI and positive with WHR), as shown in Table 2. In addition, a history of falls, physical inactivity, and difficulty with ADLs showed a positive trend with WHR but not with BMI (although there were significant differences between the quintiles of BMI in the proportions of those reporting physical inactivity and difficulty with ADLs).
|
|
Nonsmoking men and women
BMI was negatively associated with mortality risk in both men and women (P for trend = 0.0001 and < 0.0001, respectively; Table 3, model 1). These results were essentially unchanged after adjustment for potential confounders (model 2) or possible intermediate pathway variables (model 3). There was no significant trend between WC and all-cause mortality in men or women (models 1 and 2) and a negative trend in model 3. In models 1 and 2, WHR was positively but weakly associated with mortality risk in men, whereas the relation was stronger and more significantly positive in women. There was no evidence of an association with BMI and circulatory death in men, whereas there was a significant negative trend in women (Table 4). WC showed no association with circulatory mortality in men or women. There were significant positive trends with WHR and circulatory mortality in men and women (models 1 and 2), and the highest HRs were observed for the highest quintiles in both men and women. There were no significant associations with BMI or WC and cancer mortality in any models (See Supplemental Table 1
under "supplemental data" in the current online issue at www.ajcn.org). WHR showed no association with cancer mortality in men, but there were significant differences between HRs in women (P for effect = 0.024): an elevated HR was observed for the 3rd quintile compared with the lowest quintile (model 2: HR = 1.39; 95% CI: 1.02, 1.90). Deaths due to respiratory disease showed strong inverse associations for BMI (model 2; P for trend < 0.0001 in men and = 0.024 in women), whereas WC showed no association in women (P for effect = 0.66) and weak evidence for a significant inverse association in men (P for trend = 0.037). (See Supplemental Table 2 under "supplemental data" in the current online issue at www.ajcn.org.) There was a positive association with WHR and respiratory mortality in women (P for trend = 0.027) but no association in men (P for trend = 0.98). For all other causes of mortality grouped together, similar patterns of an inverse association were observed for BMI (See Supplemental Table 3 under "supplemental data" in the current online issue at www.ajcn.org). In women, WC was significantly negatively associated with mortality from other causes (model 2), and there was no relation in men. There was no association with WHR for either men or women.
|
|
). A similar result was observed for the "healthy" subset, especially the women. All analyses were consistent in finding a lower risk in all other BMI groups than in the lowest BMI group. Analyses for WHR showed no substantive change in the HRs after exclusion of the first 1 or 2 y of follow-up or in the healthy subset (Figure 2). When WHR and BMI were entered simultaneously into model 2, the positive association with WHR became stronger. In men, the HRs for higher quintiles of WHR compared with the lower quintiles were 1.23 (95% CI: 1.04, 1.44), 1.21 (1.04, 1.40), 1.34 (1.18, 1.52), and 1.47 (1.22, 1.76) (P for effect < 0.0001). In women, the comparable figures were 1.08 (95% CI: 0.97, 1.21), 1.26 (1.11, 1.42), 1.23 (1.10, 1.37), and 1.39 (1.20, 1.60) (P for effect < 0.0001). Similarly the HRs for WHR and circulatory deaths increased in magnitude after adjustment for BMI [eg, the HR for the top quintile was 1.60 in men (95% CI: 1.25, 2.04) and 1.56 (1.21, 2.01) in women].
|
|
). Those classified as underweight—ie, BMI < 18.5 (1% of men and 2.4% of women)—had consistently higher risks for all-cause mortality than did those classified as normal-weight (BMI: 18.5–24.9), but those classified as overweight (BMI: 25.0–29.9) or as obese grade 1 (BMI: 30.0–34.9) had lower rates or rates indistinguishable from those in the normal-weight group, irrespective of their WC. There was a greater risk for circulatory mortality in underweight men and men with a BMI 35–39.9 and WC > 102 cm, but few men were included in these groups. No evidence was found for any increased risks in women with very high BMIs. Conversely, there were significantly lower HRs in women classified as overweight, irrespective of WC, and in women classified as obese grade 1 with a WC > 88 cm and in men who were classified as obese grade 1 with a WC 102 cm.
|
). In women, a significant inverse trend was found for BMI and all-cause mortality, but no association was found for WHR. No significant association was found for BMI or WHR with circulatory mortality in either men or women. Deaths due to other causes are not reported because of small numbers.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
25 as being at risk, and they recommend treatment for all persons with BMI 30 or for those with BMI 25–29.9 and either 2 risk factors or a high WC (defined as 88 cm in women and 102 cm in men). Our data suggest that these recommendations are inappropriate for older (aged > 75 y) men and women, although the low prevalence of BMI 35 in older men (1.5%) limits our findings for very obese men. The prevalence of severe obesity was higher in women (4.1%), which is in line with other studies showing that severe obesity is more common in women than in men (33), but even that group did not appear to be at greater risk than were subjects with BMIs of 18.5–24.9. Our conclusions concur with those of a meta-analysis that found BMIs of 25–27 not to be a risk factor for all-cause and cardiovascular mortality in persons aged 65 y (31). Most consistently, our data highlight the risk of low BMI; persons in the lowest quintile of the BMI distribution (ie, < 23 in men and < 22.3 in women) had the highest risk of death for total mortality and for major causes of death, and men with a BMI < 18.5 were at especially high risk. These results reinforce those of a meta-analysis of follow-up studies (34) and a World Health Organization Expert Committee (35), which have also highlighted the risk of underweight. An explanation for the lack of a positive association with BMI and mortality at older ages is that, in older persons, BMI is a poor measure of body fat (36, 37). The measurement of weight does not differentiate between fat and fat-free mass, and fat-free mass (especially muscle) is progressively lost with increasing age (38). Unreliability of height measurements because of shrinkage, vertebral collapse, and measurement problems may also make BMI unreliable in older persons, but we found that substituting demi-span (ie, measurement from the sternal notch to the finger roots while the arm is outstretched laterally) for height showed patterns almost identical to those of BMI. The associations observed for all-cause mortality and BMI were also observed for circulatory, cancer, and respiratory deaths and for a heterogeneous group of other causes. Lack of specificity for an association of BMI with cause-specific mortality may reflect other factors that are more closely associated with mortality than is body fat, eg, accelerated muscle loss. In studies mainly in middle-aged persons, the U-shaped relation with BMI has been decomposed into an inverse J-shape or an inverse association with mortality and fat-free mass (5, 39-41), seen mainly in men, and a positive relation with body fat. The other main age-related change is the regional distribution of body fat with increased abdominal adiposity in older persons (42), which is reflected in increased waist (42, 43) and hip (44) circumferences in parallel with reduced subcutaneous fat stores. We found that WHR rather than WC predicted mortality in nonsmoking men and women, mainly because of the association with circulatory deaths. A few studies in older persons also examined the associations with WC or WHR and found strong positive associations with WC or WHR in men (15) or no effect of WHR but a positive association with WC and mortality (8), but only among men who never smoked and not among women. In the Health Professionals Study, in men aged > 65 y, WC and WHR were significantly related to coronary heart disease or cardiovascular disease mortality (20, 45). Stronger predictive associations for WHR than WC with mortality have been observed in young (46) and middle-aged (19) populations, whereas a number of studies in middle-aged populations have reported positive associations for both WC and WHR (20, 47-49). In a recent large case-control study, WHR was found to be more strongly associated than was BMI with myocardial infarction, whereas the association with BMI was weak and intermediate for WC (50). WC rather than WHR has also been associated with cardiovascular risk factors (51) or diabetes (52-54), principally in studies of younger populations. A limitation of WC alone as a measure is that it takes no account of body composition, whereas WHR is a measure of body shape and to some extent of lower trunk adiposity. Although it is possible theoretically for high WHR to coexist with thinness, our data show that those with high WHR had higher-than-average waist and average hip circumferences. We conclude that the association observed for WHR and mortality is probably explained by abdominal adiposity.
Our study has several limitations. The measurements were taken at a single time point, and we have no information on previous or subsequent measurements, other than a simple question about unexplained weight loss. Information on changes in weight may be important in understanding the inverse and U-shaped associations with BMI and mortality (30, 55). Our results are based on simple anthropometric indexes, and we had no direct measures of body fat or muscle composition. Recent studies have found that indexes such as midthigh circumference and muscle tissue attenuation predict functional decline (56), whereas muscle quality is a stronger predictor of mortality than is muscle quantity (57). Although we attempted to reduce measurement error through training of nurses and (for most measures) taking 2 readings, it is likely that there will be some dilution of effects due to random errors. Eligibility criteria for the trial excluded persons in nursing homes (although persons in assisted living and retirement housing were included), and the associations we describe are not generalizable to this frailer group. We did not have information on ethnicity, but, in this age group, the proportion of persons of ethnic minorities in the United Kingdom is very small. Our results may not be applicable to ethnic minorities in the United Kingdom or other populations. Although nonresponders to the study may have included those with severe obesity who were unable to attend the clinic because of limitations with mobility (58, 59), we think it unlikely that this was a reason for a nonresponse in our study, because the letter of invitation offered a home visit and one-third of participants were assessed at home.
Reverse causation is a potential problem in studies relating anthropometric indexes to subsequent mortality, but the effects are likely to be stronger in an older group with higher levels of comorbidity. In common with other authors (7, 60), we attempted to investigate the possible effects of reverse causation through excluding the early years of follow-up and restricting the analyses to a comparatively healthy subset. For BMI, this had greatest effect at the higher quintiles; it increased the HRs and brought them closer to the lowest quintile with a tendency to a U-shaped distribution, especially for circulatory causes, but in none of these analyses did the HRs for the higher BMI cross unity or did the slope become positive. There was little effect of these exclusions on the patterns observed with WHR. The follow-up time was quite short compared with some studies in middle-aged and younger groups but similar to or longer than other studies in the older group (6, 8, 12, 13, 15, 16) except one study of men aged > 70 y with a 15-y follow-up (7). Shorter duration of follow-up is inevitable in the very elderly group because of the higher death rate. Longer follow-up times also mean increasing distance from the baseline measurement, which is more likely to distort the associations in the older group with changes in body composition. Our study has several strengths compared with other studies in older persons: the sample was large, whereas other studies (7, 16, 17) may have lacked power; we used measured anthropometry rather than self-reported measurements (4, 6, 13, 19, 20); we measured several anthropometric indexes, whereas some studies measured only BMI (6, 7, 14); we were able to adjust for several confounders and associated health variables; and the large number of deaths in the current study permitted analysis by major underlying cause.
We conclude that current BMI-based health risk categories used by the World Health Organization and others to define the burden of disease related to adult overweight and obesity are not appropriate for persons aged 75 y. WHR should be used in this age group because it has a positive relation with risk of death, possibly as a result of its relation with abdominal adiposity; target cutoffs for action suggested from these analyses based on quintiles of the distribution and levels associated with the highest risks are WHR > 0.99 in elderly nonsmoking men and > 0.90 in elderly nonsmoking women. Our conclusions for the group aged 75 y support recent recommendations to use WHR rather than BMI in middle-aged persons (50, 61). We also recommend that attention be paid to the problem of underweight in old age. Men and women whose BMI values are < 23 and < 22.3, respectively, require further investigation, including assessment of diet and health status.
|
ACKNOWLEDGMENTS |
|---|
AF is the principal investigator and CB is the co-investigator of the MRC Trial of the Assessment and Management of Older People in the Community; they designed and implemented the trial. EB helped to administer the study and train the nurses and edited data. RU contributed to the study design. GP and AF conducted the statistical analysis and drafted the manuscript. All authors have commented on drafts of the manuscript. AF and GP had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. None of the authors had a personal or financial conflict of interest.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  T. W. Auyeung, J. S. W. Lee, J. Leung, T. Kwok, P. C. Leung, and J. Woo Survival in Older Men May Benefit From Being Slightly Overweight and Centrally Obese--A 5-Year Follow-up Study in 4,000 Older Adults Using DXA J Gerontol A Biol Sci Med Sci, July 23, 2009; (2009) glp099v1. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 E. V. Sanchez-Vaznaugh, I. Kawachi, S. V. Subramanian, B. N. Sanchez, and D. Acevedo-Garcia Do Socioeconomic Gradients in Body Mass Index Vary by Race/Ethnicity, Gender, and Birthplace? Am. J. Epidemiol., May 1, 2009; 169(9): 1102 - 1112. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. M Flegal and B. I Graubard Estimates of excess deaths associated with body mass index and other anthropometric variables Am. J. Clinical Nutrition, April 1, 2009; 89(4): 1213 - 1219. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. E. Litwin Which Measures of Obesity Best Predict Cardiovascular Risk? J. Am. Coll. Cardiol., August 19, 2008; 52(8): 616 - 619. [Full Text] [PDF]
|
|
|
|
|
|
A. Koster, M. F. Leitzmann, A. Schatzkin, T. Mouw, K. F. Adams, J. Th. M. van Eijk, A. R. Hollenbeck, and T. B. Harris Waist Circumference and Mortality Am. J. Epidemiol., June 15, 2008; 167(12): 1465 - 1475. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 V. Hirani and J. Mindell A comparison of measured height and demi-span equivalent height in the assessment of body mass index among people aged 65 years and over in England Age Ageing, May 1, 2008; 37(3): 311 - 317. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. J. Cameron and P. Z. Zimmet Expanding Evidence for the Multiple Dangers of Epidemic Abdominal Obesity Circulation, April 1, 2008; 117(13): 1624 - 1626. [Full Text] [PDF]
|
|
|
|
|
|
C. Zhang, K. M. Rexrode, R. M. van Dam, T. Y. Li, and F. B. Hu Abdominal Obesity and the Risk of All-Cause, Cardiovascular, and Cancer Mortality: Sixteen Years of Follow-Up in US Women Circulation, April 1, 2008; 117(13): 1658 - 1667. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S G. Wannamethee, A G. Shaper, L. Lennon, and P. H Whincup Decreased muscle mass and increased central adiposity are independently related to mortality in older men Am. J. Clinical Nutrition, November 1, 2007; 86(5): 1339 - 1346. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P. Poirier Adiposity and cardiovascular disease: are we using the right definition of obesity? Eur. Heart J., September 1, 2007; 28(17): 2047 - 2048. [Full Text] [PDF]
|
|
|
|
|
|
E. Cereda, V. Sansone, G. Meola, and A. E. Malavazos Increased visceral adipose tissue rather than BMI as a risk factor for dementia Age Ageing, September 1, 2007; 36(5): 488 - 491. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L. Sjostrom, K. Narbro, C. D. Sjostrom, K. Karason, B. Larsson, H. Wedel, T. Lystig, M. Sullivan, C. Bouchard, B. Carlsson, et al. Effects of Bariatric Surgery on Mortality in Swedish Obese Subjects N. Engl. J. Med., August 23, 2007; 357(8): 741 - 752. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L. Ferrucci and D. Alley Obesity, Disability, and Mortality: A Puzzling Link Arch Intern Med, April 23, 2007; 167(8): 750 - 751. [Full Text] [PDF]
|
|
|
|
 |
|
 J. Lindsay Research Batters BMI's Utility DOC News, November 1, 2006; 3(11): 6 - 6. [Full Text]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
, Main analyses (all included); 
, first 12 mo of follow-up excluded; 
, first 24 mo of follow-up excluded; and 
, "healthy" persons only and all follow-up included. "Healthy" was defined as no history at assessment of cancer, cardiovascular disease, diabetes, Parkinson disease, hip fracture, angina, weight loss, pneumonia, or leg ulcer; not cognitively impaired; able to carry out 