Body Weight and Weight Change and Their Health Implications for the Elderly

Body weight and weight change and their health implications for the elderly

JC Seidell1,2* and TLS Visscher1,3

1Department of Chronic Diseases Epidemiology, National Institute of Public Health, Bilthoven, The Netherlands; 2Institute for Research in Extramural Medicine, Free University Amsterdam, The Netherlands; and 3Netherlands Institute for Health Sciences, Erasmus University, Rotterdam, The Netherlands

After the age of 60 y, body weight on average tends to decrease. The contribution of fat mass to this weight loss is relatively small, but fat tends to be redistributed with advancing age toward more abdominal (particularly visceral) fat. Anthropometric data are relatively poor indicators of these aging processes. This may be one of the explanations why the relationship between high body mass index and mortality is less pronounced in older than in younger people. Reduced lipolysis in the visceral fat depot with aging is among potential explanations why increased visceral fat seems to be less harmful in elderly subjects compared to young adults. Even though the relative contribution of increased fat mass to mortality may be less pronounced in elderly people, the impact on disability and functional limitations is found to be important from both a clinical and a public health point of view. At the other end of the scale studies have shown that low body mass index and weight loss in the elderly are both strong predictors of subsequent mortality. This cannot be explained by effects of smoking and early mortality after baseline. There are only few systematic studies comparing the predictive validity of different anthropometric data for mortality. One recent prospective study showed that a high waist circumference (in non- smoking men) may be a better predictor of all-cause mortality than high body mass index and waist=hip ratio. Low BMI was a better predictor of mortality than low waist circumference. In conclusion changes in body composition and fat distribution with aging are poorly captured by standard anthropometric data. Low lean body mass is probably better re¯ected by low BMI, whereas increased (abdominal) fatness is better re¯ected by increased waist circumference. Descriptors: aging; obesity; fatness; mortality; morbidity; fat distribution European Journal of Clinical Nutrition (2000) 54, Suppl 3, S33±S39

Epidemiology of weight status and aging Effect of age on body composition and body fat distribution In cross-sectional studies the prevalence of high body weight or obesity (body mass index (BMI) > 30 kg=m2) One fundamental problem in the interpretation of such increases with age up to about age 60 and then declines (see epidemiological data is that anthropometric data have Figure 1). different implications for body composition at different There can be several explanations for this: ages. For instance from densitometry (Deurenberg et al, 1991) it has been calculated that selective survival (obese young and middle-aged persons have died prematurely); percentage body fat cohort-effect (old people come from cohorts in which 1:20 Â BMI 0:23 Â Age 10:8 obesity was less common); 2 people start to loose weight after age 60 y (Shock, ÂSex men 1; women 05:4 r 0:79 1972). This implies for instance that at a BMI of 30 kg=m2 a man All three possibilities are likely to play some role, but aged 20 y has 24.4% body fat, whereas a man of 80 y has their exact quantitative contribution has not been studied 38.8% body fat and that women have a much larger relative systematically yet. There are some changes associated with fat mass than men at every age and level of BMI. In aging which could actually be expected to have an effect in addition, body fat is redistributed with increasing age in the opposite direction, such as a decline in stature which, the sense that more of it becomes located in the abdominal even when weight is maintained in all people, would lead to cavity (visceral fat) (Borkan et al, 1985; Seidell et al, 1988; an increase in the prevalence of obesity with aging (van Figure 2). The reason for this redistribution is not really Leer et al, 1992; Launer et al, 1995). known, but declining testosterone and growth hormone levels in combination with declining rates of lipolysis of visceral fat with aging (Ostman et al, 1969) may play a role in men. Low testosterone levels have been shown to be *Correspondence: JC Seidell, Department of Chronic Diseases Epidemiology, National Institute of Public Health and the Environment, associated with increased visceral fat mass in men (Seidell PO Box 1, 3720 BA Bilthoven, The Netherlands. et al, 1990). In women the declining oestrogen levels after E-mail: [email protected] menopause may be a critical factor. Body weight and weight change JC Seidell and TLS Visscher S34

Figure 1 Prevalence of obesity by age in men and women in the United Kingdom 1997 (Health Survey for England 1997).

Figure 2 Visceral fat area as a percentage of total fat area in the abdomen assessed by computed tomography by age in men and women (adapted from Seidell et al, 1988).

In addition, weight loss at older age does not re¯ect the nounced with aging (Andres et al, 1985; Stevens et al, same changes as at younger age. In the elderly, it is much 1998, Waaler, 1988; see Figure 3). more closely associated with loss of lean body mass than in There are several potential explanations for this: young adults. This accelerated loss of fat-free mass in older people is often called sarcopenia (Poehlman et al, 1995). 1. BMI is not an optimal indicator of body composition in Involuntary weight loss is frequently reported in elderly the elderly (Visser et al, 1994). Allison et al (1997) patients and usually caused by acute or chronic diseases speculated that the U-shaped relation between BMI and (Fischer & Johnson, 1990). mortality resulted from an inverse linear association In conclusion, although the prevalence of obesity between lean body mass and mortality and a positive declines with age after age 60, this does not imply that association between body fat mass and mortality. The excess fat storage is uncommon in the elderly; progres- BMI does not allow separation of these two phenomena sively more fat is stored in the abdominal cavity. Anthro- (Allison et al, 1997). pometric data such as weight, height and skinfold 2. Selective survival: obese persons at high risk have died thicknesses seem inadequate to measure these changes. prematurely and what remains is a selection of relatively healthy obese subjects. 3. Ceiling effect: absolute mortality rates increase with age and for instance at 90 y of age 15 y mortality will be Effect of age on the association between BMI and health close to 100% regardless of any risk factor status. 4. Cohort effect: obese subjects now at an old age have Mortality been exposed to lifestyles and environments that are Many studies have documented that the U-shaped relation different from those of future obese elderly persons. between BMI and all-cause mortality becomes less pro- For instance someone now aged 85 y was born at a time

European Journal of Clinical Nutrition Body weight and weight change JC Seidell and TLS Visscher S35

Figure 3 Ten-year mortality in Norwegian men by age and BMI (adapted from Waaler, 1988). Lower line: age 45 ± 49; middle line: age 60 ± 64; upper line: age 75 ± 79.

when life expectancy at birth was about 50 y and the instance, obesity has very little effect on total mortality at most frequent cause of death was infectious disease body mass indices below 40 kg=m2 (Pettitt et al, 1982). whereas obesity and co-morbidities such as heart disease were relatively uncommon. Visscher et al attempted to Morbidity separate age and cohort effects in aging men and Much of the available data on BMI and health are restricted showed that when a cohort effect is excluded age has to mortality. There is, however, increasing evidence that an little effect on the relative risks associated with high excessive BMI in elderly women and men is associated BMI (Visscher et al, 1999). with impairment of health, disability and reduced quality of 5. Excess fat is less detrimental in older than in younger life in terms of functional limitations (Launer et al, 1994; people. This issue will be explored further in the section Visser et al 1998a, 1998b). In several studies Visser et al on body fat distribution. (1998a, 1998b) explored the possibility that disability is Although the relative risks of a high BMI may become caused by low lean body mass and high fat mass. In all less pronounced with aging, the relatively high prevalence instances they concluded that particularly high body fat of obesity in the younger elderly, in combination with high mass is associated with mobility disability and general mortality rates (absolute risks), make a high BMI an disability even in the very old (Figures 4 ± 6). important public health issue. Comparative data on these risks in young and elderly are At the other end of the scale studies have shown that low not available, but given the large incidence of functional BMI and weight loss in the elderly are both strong pre- limitations and disability in the elderly this is certainly dictors of subsequent mortality. This cannot be explained becoming a more important public health issue with advan- by effects of smoking and early mortality after baseline cing age. (Visscher et al, 1999). There has been some debate as to whether or not the Effect of age on the association between weight change ¯attening of the BMI:mortality curve with aging and the and health upward shift of the nadir (the point of minimum mortality) of the curve with age should lead to the conclusion that cut- On average people above about 60 y of age lose weight off points for obesity should be higher in older than in when aging. Many observational studies suggest that younger people (Andres et al, 1985). Such implications are weight changes (weight increase, weight loss and weight very dif®cult to derive from cross-sectional analyses. Par- ¯uctuations) are predictors of mortality (Peters et al, 1995). ticularly the implication that minimum mortality can only In the elderly, most weight loss may be involuntary and be maintained with continued weight gain during adult life caused by disease. Psycho-social factors may also play an has been criticised, especially since adult weight gain is an important role. There is no information on the effects of independent predictor of mortality and morbidity (Willett intentional weight loss in the elderly, and the increased et al, 1991). mortality and morbidity in elderly subjects who lose weight Almost all studies on BMI and mortality and aging have may thus be spurious. been carried out in Caucasian subjects living in the US or Weight loss in elderly subjects is likely to re¯ect mainly Europe, and the relationship remains relatively unexplored a loss of lean body mass (particularly muscle) and this may in other ethnic groups. In younger subjects there are contribute to the increased mortality, especially if the initial indications that the relationship between BMI and mortality body weight is relatively low. Weight loss has been found is similar in some ethnic groups such as the Japanese to be associated with increased mortality, whereas in the (Tsukamoto & Sano, 1990) and may also be very different same studies body fat loss was associated with reduced from the relationship in Caucasians. In Pima Indians, for mortality (Allison et al, 1999).

European Journal of Clinical Nutrition Body weight and weight change JC Seidell and TLS Visscher S36

Figure 4 Relative risk (RR) for the onset of mobility disability in American White women by tertiles of past BMI (measured on average 14 y prior to the assessment of disability). Asterisks indicate signi®cantly elevated relative risk compared with the lowest BMI category. Young-old women were 45 ± 59 y at baseline; old-old women were 60 ± 74 y at baseline. Relative risks adjusted for age, smoking status, educational level and time-to follow-up. Adapted from Launer et al (1994).

Figure 5 Relative risk (RR) for the onset of mobility disability in American White women by weight change during ®rst 5 y of follow-up (baseline weight measured on average 14 y prior to the assessment of disability). Asterisks indicate signi®cantly elevated relative risk compared to the weight stable category. Young-old women were 45 ± 59 y at baseline; old-old women were 60 ± 74 y at baseline. Relative risks adjusted for age, smoking status, educational level and time-to follow-up. Adapted from Launer et al (1994).

Weight loss may obscure true associations between BMI co-morbidities may respond favourably to weight loss in and morbidity in old age. Harris et al (1997) showed that a those who are overweight. high BMI was associated with increased coronary heart disease in late middle age, but not in old age. This difference in risk was due to weight loss between middle Effect of age on the association between fat distribution and old age. High BMI remained a risk factor for coronary and health heart disease once those with substantial weight loss were Because of the redistribution of body fat with aging, BMI excluded. In elderly subjects with a BMI in the acceptable becomes a poor indicator of both overall fatness and range, weight stability is of great importance. Because of abdominal fatness. Waist=hip circumference ratio and, the lack of results of intervention studies it is hazardous to more recently, waist circumference alone have been recom- recommend intentional weight loss in the elderly, even mended as better indicators of abdominal fatness than BMI. when they are overweight, although in theory many of the Waist circumference may be a better indicator than BMI of

European Journal of Clinical Nutrition Body weight and weight change JC Seidell and TLS Visscher S37

Figure 6 Relative odds (error bars denote 95% con®dence intervals) for mobility disability in men and women aged 75 ± 95 y by tertiles of percent total body fat, adjusted for age, education, self-rated health, chronic illness, physical activity, estrogen use (women only), alcohol use, and smoking. Adapted from Visser et al (1998a). overall fatness as well. Waist=hip ratio in the elderly may comparing Indians living in Sweden with Swedish men be dif®cult to interpret because, whereas the waist mea- support this hypothesis (Chowdhury et al, 1996). Indian sures abdominal fatness, hip circumference may re¯ect also men with the same height and weight have higher risk variation in pelvic width and gluteal muscle. A recent factors (blood pressure, insulin and triglycerides) and cross-sectional study comparing anthropometric data of waist=hip ratios than Swedish men. Whole body computed patients with type 2 diabetes mellitus with those of healthy tomography revealed that there were no differences in body controls, suggests that the association between high waist= fat mass and fat distribution between the two groups; the hip ratio and diabetes may be due to both relatively large only signi®cant difference was the relatively low leg waist and narrow hip circumferences, especially in men muscle mass in the Indians. This aspect of peripheral (Seidell et al, 1997). One hypothesis is that narrow hips muscle wasting is likely to be much more pronounced in re¯ect peripheral muscle wasting and that this is an impor- elderly subjects, and the interpretation of waist=hip ratio tant correlate of diabetes mellitus. More detailed studies may be increasingly dif®cult with advancing age.

Figure 7 Glycerol release (lipolytic response) to noradrenaline from omental adipose tissue by age (adapted from Ostman et al, 1969). Error bars represent standard deviation.

European Journal of Clinical Nutrition Body weight and weight change JC Seidell and TLS Visscher S38 There have been few systematic comparisons in elderly absence of body composition data, the waist circumference subjects of the relation between BMI, waist circumference may be more revealing than BMI. and waist=hip ratio on the one hand and health on the other. Visscher et al (1999) recently showed that in men aged 55 or more a large waist circumference was a better predictor of higher 5 y mortality than waist=hip ratio or BMI. A References positive association between waist circumference and mortality was even seen in men aged 70 y or older (Visscher Allison DB, Faith MS, Heo M & Kotler DP (1997): Hypothesis concerning et al, 2000). Low BMI was a better predictor of mortality the U-shaped relation between body mass index and mortality. Am. J. Epidemiol. 146, 339 ± 349. than low waist circumference. It could be that low lean Allison DB, Zannolli R, Faith MS, Heo M, Pietrobelli A, van Itallie TB, body mass is better re¯ected by low BMI whereas PI-Sunyer FX & Heyms®eld SB (1999): Weight loss increases and fat increased (abdominal) fatness is better re¯ected by loss decreases all-cause mortality rate: results from two independent increased waist circumference. cohort studies. Int. J. Obes. Relat. Metab. Disord. 23, 603 ± 611. Andres R, Elahi D, Tobin JD, Muller DC & Brant L (1985): Impact of age Comparative data on the impact of abdominal fatness on on weight goals. Ann. Intern. Med. 103, 1030 ± 3. mortality in younger and older subjects is scarce. Results of BjoÈrntorp P (1990): `Portal' adipose tissue as a generator of risk factors for the Baltimore Longitudinal Study on Aging show that cardiovascular disease and diabetes. Arteriosclerosis 10, 493 ± 496. increased abdominal depth (sagittal diameter) was related Borkan GA, Hults DE, Gerzof SG & Robbins AH (1985): Comparison of to increased all-cause mortality and coronary heart disease body composition in middle-aged and elderly males using computed tomography. Am. J. Anthropol. 66, 289 ± 295. (CHD) mortality in men aged 55 y or younger, whereas this Chowdhury B, Lantz H & SjoÈstroÈm L (1996): Computed tomography- same diameter was not related to either endpoint in older determined body composition in relation to cardiovascular risk factors men (Seidell et al, 1994). Again, explanations include in Indian and matched Swedish males. Metabolism 45, 634 ± 644. selective survival and cohort effects, but it may also be Deurenberg P, Weststrate JA & Seidell JC (1991): Body mass index as a measure of body fatness: age- and sex- speci®c prediction formulas. Br. that the increased abdominal accumulation of fat is less J. Nutr. 65, 105 ± 114. hazardous in older than in younger men. One well-accepted Fischer J & Johnson MA (1990): Low body weight and weight loss in the theory linking intra-abdominal or visceral fat to health is aged. J. Am. Diet. Assoc. 90, 1697 ± 1706. the release of free fatty acids into the portal vein thereby Frayn KN, Samra JS & Summers LK (1997): Visceral fat in relation to exposing the liver to high concentrations of free fatty acids health: is it a major culprit or simply an innocent bystander? Int. J. Obes. Relat. Metab. Disord. 21, 1191 ± 1192. (BjoÈrntorp, 1990). Although there are many potential con- Harris TB, Launer LJ, Madans J & Feldman JJ (1997): Cohort study of founding factors (Seidell & Bouchard, 1997) and excep- effect of being overweight and change in weight on risk of coronary tionally high concentrations of free fatty acids in the portal heart disease in old age. BMJ 314, 1791 ± 1794. vein have not been demonstrated in humans in vivo (Frayn Launer LJ, Harris T, Rumpel C & Madans J (1994): Body mass index, weight change, and risk of mobility disability in middle-aged and older et al, 1997) this is still an attractive hypothesis. Ostman et women. The epidemiologic follow-up study of NHANES I. JAMA 271, al (1969) showed that stimulated lipolysis in omental fat is 1093 ± 1098. progressively reduced with aging, and this may mean that Launer LJ, Barendregt JJ & Harris T (1995): Shift in body mass index fat accumulated in the abdomen is a less important source distributions due to height loss. Epidemiology 6, 98 ± 99. of free fatty acids in the elderly than in younger adults Ostman J, Efendic S & Arner P (1969): Catecholamines and metabolism of human adipose tissue. I. Comparison between in vitro effects of (Figure 7). noradrenaline, adrenaline and theophylline on lipolysis in omental adipose tissue. Acta Med. Scand. 186, 241 ± 246. Peters ET, Seidell JC, Menotti A, Aravanis C, Dontas A, Fidanza F, Conclusions Karvonen M, Nedeljkovic S, Nissinen A, Buzina R, Bloemberg B & Kromhout D (1995): Changes in body weight in relation to mortality in In subjects aged 60 y or older rates of obesity decrease, but 6441 European middle-aged men: the Seven Countries Study. Int. J. this may not adequately re¯ect increased fatness and Obes. Relat. Metab. Disord. 19, 862 ± 868. abdominal fat accumulation with aging. Conventional Pettitt DJ, Lisse JR, Knowler WC & Bennett PH (1982): Mortality as a anthropometric indicators such as BMI and waist=hip function of obesity and diabetes mellitus. Am. J. Epidemiol. 115, 359 ± 366. ratio may not be adequate measures of overall and abdom- Poehlman ET, Toth MJ, Fishman PS, Vaitkevicius P, Gottlieb SS, Fisher inal fatness in elderly subjects. Waist circumference may ML & Fonong T (1995): Sarcopenia in aging humans: the impact of be a better indicator of body composition, fat distribution menopause and disease. J. Gerontol. A. Biol. Sci. Med. Sci. 50(Special and impaired health. issue), 73 ± 77. Reduced lipolysis may be an explanation why increased Seidell JC & Bouchard C (1997): Visceral fat in relation to health: is it a major culprit or simply and innocent bystander? Int. J. Obes. Relat. (abdominal) fat is associated with lower relative risks in Metab. Disord. 21, 626 ± 631. older than in younger people. Because of its high preva- Seidell JC, Oosterlee A, Deurenberg P, Hautvast JGAJ & Ruijs JHJ lence in older people and the high mortality and morbidity (1988): Abdominal fat depots measured with computed tomography: rates, increased (abdominal) fatness may nevertheless effects of degree of obesity, sex and age. Eur. J. Clin. Nutr. 42, 805 ± 815. remain an important public health problem. Seidell JC, BjoÈrntorp P, SjoÈstroÈm L, Kvist H & Sannerstedt R (1990): Weight stability within the acceptable range or even Visceral fat accumulation in men is positively associated with insulin, slightly above seems to result in the lowest mortality in the glucose, and C-peptide levels, but negatively with testosterone levels. elderly. However, weight loss with aging particularly at Metabolism 39, 897 ± 901. low BMI may be partly responsible for the seemingly high Seidell JC, Andres R, Sorkin JD & Muller DC (1994): The sagittal waist diameter and mortality in men: the Baltimore Longitudinal Study on optimal BMI. In addition, high BMI (especially high body Aging. Int. J. Obes. Relat. Metab. Disord. 18, 61 ± 67. fat mass) is associated with increased risk of disability. It is Seidell JC, Han TS, Feskens EJM & Lean MEJ (1997): Narrow hips and probably inappropriate to tolerate high levels of fatness in broad waist circumferences independently contribute to increased risk old people because of the relationship between BMI and of non-insulin-dependent diabetes mellitus. J. Int. Med. 242, 401 ± 406. Shock NW (1972): Energy metabolism, caloric intake and physical activity mortality. There is a greater need for body composition of the aging. In: Nutrition in Old Age, X Symposium Swedish Nutrition data rather than only measurements of body weight both in Foundation ed. LA Carlson, pp 372 ± 383. Uppsala: Almqvist & clinical situations and in epidemiological studies. In the Wiksell.

European Journal of Clinical Nutrition Body weight and weight change JC Seidell and TLS Visscher S39 Stevens J, Cai J, Williamson DF, Thun MJ & Wood JL (1998): The effect Visser M, Harris TB, Langlois J, Hannan MT, Roubenoff R, Felson of age on the association between body mass index and mortality. N. DT, Wilson PWF & Kiel DP (1998a): Body fat and skeletal Engl. J. Med. 338,1±7. muscle mass in relation to physical disability in very old men Tsukamoto H & Sano F (1990): Body weight and longevity; insurance and women of the Framingham Heart Study. J. Gerontol. 53A, experience in Japan. Diabetes Res. Clin. Pract. 10(Suppl 1), S119 ± S125. M214 ± 221. Van Leer EM, van Noord PAH & Seidell JC (1992): Components of adult Visser M, Langlois J, Guralnik JM, Kronmal RA, Robbins J, height and height loss Ð secular trend and effects of aging in women in Williamson JD & Harris TB (1998b): High body fatness, but the DOM project. Ann. Epidemiol. 2, 611 ± 615. not low fat-free mass, predicts disability in older men and Visscher TLS, Seidell JC, Menotti A, Blackburn H, Nissinen A, Feskens women: the Cardiovascular Health Study. Am. J. Clin. Nutr. 68, EJM & Kromhout D (2000): Under- and overweight in relation to 584 ± 590. mortality among men aged 40 ± 59 and 50 ± 69: the seven countries Waaler H (1988): Hazard of obesity Ð the Norwegian experience. Acta study. Am. J. Epidemiol. (in press). Med. Scand. 723(Suppl), 17 ± 21. Visser M, van den Heuvel E & Deurenberg P (1994): Prediction equations Willett WC, Stampfer M, Manson J & Vanltallie T (1991): New weight for the estimation of body composition in the elderly using anthropo- guidelines for Americans: justi®ed or injudicious? Am. J. Clin. Nutr. 53, metric data. Br J. Nutr. 71, 823 ± 833. 1102 ± 1103.

European Journal of Clinical Nutrition