Thesis. Reference. Body composition : methods of measurement, normative values and clinical use. GENTON GRAF, Laurence

Transcription

1 Thesis Body composition : methods of measurement, normative values and clinical use GENTON GRAF, Laurence Abstract Measurement of body composition is an important part of nutritional assessment. The low FFM associated with malnutrition has been associated with numerous infectious and noninfectious complications, increasing length of stay, morbidity and mortality. DXA and TBK are reference method for determination of FFM and BCM, but, as these methods are expensive and require extensive technique of the operator, we have focused especially on BIA, an easy, quick, safe and reliable bedside method to measure body composition. BIA formulas to routinely assess FFM and appendicular skeletal muscle mass have been developed. Normative values of total body composition have been established, according to age and gender. Longitudinal and cross-sectional studies allowed an insight on the impact of physical activity and environment on body composition. With regard to clinics, we have studied the impact of body composition, determined by BIA, on length of hospital stay and shown that a low FFM and FM index were associated with an increased length of stay. This demonstrates that prevention of FFM loss, whether through nutritional support or drugs, [...] Reference GENTON GRAF, Laurence. Body composition : methods of measurement, normative values and clinical use. Thèse de privat-docent : Univ. Genève, 2011 DOI : /archive-ouverte/unige:17006 Available at: Disclaimer: layout of this document may differ from the published version.

2 BODY COMPOSITION: METHODS OF MEASUREMENT, NORMATIVE VALUES AND CLINICAL USE Laurence Genton, MD 2011 Address : Dr. Laurence Genton Service d Endocrinologie, diabétologie et nutrition Département de médecine interne Hôpitaux Universitaires de Genève 21, rue Micheli-du-Crest 1211 Genève 14 Tel/Fax : / laurence.genton@hcuge.ch Thèse de privat-docent Faculté de Médecine Université de Genève

3 BODY COMPOSITION: METHODS OF MEASUREMENT, NORMATIVE VALUES AND CLINICAL USE Laurence Genton, MD 2011 Address : Dr. Laurence Genton Service d Endocrinologie, diabétologie et nutrition Département de médecine interne Hôpitaux Universitaires de Genève 21, rue Micheli-du-Crest 1211 Genève 14 Tel/Fax : / laurence.genton@hcuge.ch Thèse de privat-docent Faculté de Médecine Université de Genève 1

4 CONTENT 1. Introduction Definition of body composition Importance of body composition determination Health implications of body composition changes Factors influencing body composition in healthy subjects Oral intakes Other factors 8 2. Methods for body composition measurements Reference methods Generalities Total body potassium Dual-energy X-ray absorptiometry Bedside methods Generalities Skinfold thickness and circumferences Bioelectrical impedance analysis Normative values of total body composition Swiss normative values Impact of aging Impact of physical activity Impact of obesity Impact of environment Clinical use of total body composition Nutritional risk Nutritional follow-up Summary and conclusion Acknowledgements References 34 2

5 1. INTRODUCTION Definition of body composition Since the beginning of the century, many studies have tried to assess body composition. However, body composition had not been clearly defined, and led to overlapping or omission of components in many studies. In order to clarify confusing terminology, Wang et al. have defined the components of body composition by five levels of increasing complexity (table 1) (1). Table 1: The five levels of body composition (adapted from (1)) I II III IV V Atomic Molecular Cellular Tissue-system Whole-body Carbon Water Extracellular Blood Hydrogen Protein solid Bone Oxygen Lipid Extracellular Fat mass Nitrogen Glycogen fluid Skeletal muscle Calcium Minerals Body cell mass Other Presently, the most convenient approach to body composition assessment is to divide the human body into two components, fat (FM) and fat-free mass (FFM) (2). This model relies on the molecular level, where water, protein, glycogen, mineral and essential lipids build up FFM and non-essential lipids correspond to fat mass. Essential lipids, as sphingomyelin and phospholipids, are found in the bone marrow, heart, lungs, liver, spleen, kidneys, intestines, muscles and central nervous system. They serve physiological functions as forming cell membranes but it is not clear whether they can be used as metabolic fuel. In contrast, non-essential lipids, largely triglycerides, accumulate in visceral and subcutaneous fat. They protect organs from trauma and thermal stress and serve as energy storage. 3

6 Stricto sensu, lean body mass is not synonymous of FFM as it contains both essential and non-essential lipids. However, in the literature, they are often used interchangeably Importance of body composition determination Determination of body composition is part of the nutritional assessment, as well as clinical history, medical examination, anthropometrics and biological markers (table 2). Table 2: Parameters allowing assessment of nutritional state (adapted from (3)) Parameters Clinical history Medical examination Weight, Height Biological markers Body composition Energy balance (needs vs. intakes and losses), weight history, appetite Abnormalities of skin, hair, nails, digestive symptoms, edema Calculation of body mass index Plasma: total protein, albumin, transthyretin, transferrin, IGF-1, lymphocyte count Urine: creatine, 3-méthyl-histidine, Skinfold thickness, circumferences, bioelectrical impedance analysis, dual-energy x-ray absorptiometry However, body composition is often not measured because of unavailability of measurement methods, and lack of experience and knowledge. consequently, nutritional assessment often relies only on anthropometry, laboratory values and evaluation of energy balance (4). Unfortunately, anthropometric parameters give only a crude estimation of body composition. In women, a high body mass index (BMI = weight (kg)/height (m) 2 )) has been associated with a high FFM and FM while in men it may reflect only high FM (5). In patients, as those with chronic hypercapnic failure, BMI may underestimate FFM depletion (6). Knowing that changes in FFM and FM carry health implications, assessment of body composition is preferable to anthropometrics for guidance of nutritional care. 4

7 1. 3. Health implications of body composition changes Changes in absolute FFM and FM carry health implications. A decrease of FFM is a hallmark of protein energy malnutrition (PEM). The literature describes several types of PEM, i.e. cachexia, starvation and/or sarcopenia. Cachexia represents a complex metabolic syndrome associated with underlying illness, characterized by a weight loss of at least 5% in 12 months and three of the following criteria: low FFM, decreased muscle strength, fatigue, anorexia, and abnormal biological markers (CRP>5.0 mg/l, hemoglobin < 120g/l, serum albumin <33 g/l) (7). In contrast, starvation results from a pure deficit of all macro- and micronutrients, as seen for instance in hunger strikers and persons with anorexia. Sarcopenia describes the loss of skeletal muscle mass and function occurring mostly in older or immobilized subjects but it is not quite sure whether it reflects a third type of PEM, as it is associated with increased plasma concentrations of inflammatory cytokines. The physiological mechanisms of decreased FFM are summarized on figure 1. Figure 1: The physiological mechanisms of decreased FFM (adapted from (8, 9)) - Energy and protein intakes - Insulin - Anabolic hormones - Pro-inflammatory cytokines - Anabolic hormones - Starvation - Anorexia - Hypermetabolic state Cachexia - Neurodegenerative process - Muscle fiber atrophy - Disability/inactivity - Testosterone, estrogens, growth hormone, insulin-like-growth factor-1, 25-hydroxy ergocalciferol, dehydroepiandraosterone - or imbalance in protein metabolism - Basal metabolic rate - Genetic factors Starvation Loss of fat-free mass Sarcopenia 5

8 The loss of FFM occurs mainly in peripheral skeletal muscle mass. It induces a loss of muscle strength, a high risk of disability, balance disorders and falls, and an impaired quality of life (10). It also affects respiratory muscle mass, translating clinically into impaired lung function and exercise tolerance (11, 12). The other consequences of PEM and associated decrease in FFM are delayed wound healing and immune dysfunction, which favor infection and ultimately promote further catabolism (13). Finally, there is evidence that a low FFM prolongs the ICU and hospital stays (14), increases morbidity and mortality (15), delays rehabilitation and increases therapeutic costs (16, 17). In contrast, a high FM is related to overweight and obesity. The risks associated with obesity have been described by the World Health Organization (table 3) (18). Table 3: Relative risk of health problems associated with obesity (adapated from (18)) Relative risk Health problems > 3 Insulin resistance, non-insulin dependent diabetes, cholecystopathy, dyspnea, sleep apnea 2-3 Coronary heart disease, hypertension, knee osteoarthritis, hyperuricemia and gout 1-2 Cancer (breast, uterus, colon), polycystic ovarian syndrome, abnormalities in fertility, fetal abnormalities, lumbar pain increased risk of anaesthesia complications The accumulation of abdominal FM is especially detrimental for health. High abdominal FM plays an important role in the appearance of insulin resistance and the metabolic syndrome (18), as well as in the appearance of functional limitations and disability (19). Altogether, neither the loss of FFM nor the gain of FM is beneficial for overall health. 6

9 1. 4. Factors influencing body composition in healthy subjects Several factors influence body composition in healthy subjects. These are mainly age, gender, physical activity or physical exercise, and oral intakes. Their understanding is essential in order to interpret adequately changes of body composition occurring during disease. We have studied especially the former three parameters and their impact on body composition will be detailed in chapter 3. This chapter summarizes the impact of oral intakes and other factors on body composition Oral intakes The impact of energy restriction on body composition has been studied mostly in overweight or obese subjects, whether in association or not with physical exercise. Physical exercise corresponds to any physical activity that enhances or maintains physical fitness and overall health and wellness. Several meta-analysis have shown that energy restriction reduces body weight by FM and FFM loss (20-22). Non-exercising dieters loose approximately 25% of their weight loss as FFM and 75% as FM (21, 22). According to a recent systematic review, modest weight loss generates preferential loss of visceral FM rather than subcutaneous FM but greater weight loss attenuates this effect (23). It remains however unknown whether the changes of regional FM during diet are gender-dependent. The addition of physical exercise to calorie restriction limits the loss of FFM. A metaanalysis focused on studies with at least one exercising and one non-exercising group. It found that, for a weight loss of 10 kg, the loss of FFM was 1.7 kg with an intervention combining diet plus exercise compared to 2.5 kg with energy restriction alone (21). The impact of calorie supplementation has not been studied extensively in healthy subjects but leads to weight gain. In an old study, male non obese volunteers with an initial FM of 15% increased their calorie intake to about 7000 kcal/d for 40 weeks. Their body mass 7

10 increased by 25% and percent FM doubled (24). Forbes differentiates the changes in body composition according to initial body mass. In case of weight gain, he describes a 60-70% increase of FFM in thin people compared to 30-40% of FFM in the obese subjects (25). When combined with physical exercise, especially resistance training, calorie supplementation improves the proportion of weight gained as FFM. Rozenek et al. randomized 73 non obese men to a calorie supplement of 2010 kcal/d or no supplement for 8 weeks in addition to their standard diet (26). All subjects were submitted to a resistance training 4 d/wk. Intakes, calculated at week 1, 4 and 8 by 3-day dietary histories, were 1700 kcal/d higher in the supplemented group compared to the control group. Within 8 weeks, the supplemented group increased weight by 3 kg or 100g/d consisting almost exclusively of FFM. Old subjects can also improve their muscle mass and function with energy supplementation and simultaneous resistance training (27, 28). It remains unclear whether these benefits come from the extra energy or specifically from the increased protein intakes. Thus, hypercaloric diets may increase FFM when resistance exercise is performed simultaneously. In a recent review, we have summarized the impact of macronutrient intakes and physical exercise on physical fitness, including body composition (29) Other factors Other factors like race, smoking, educational level, parity and menopause influence body weight and composition and its distribution. Many of the studies dealing with these factors relied on anthropometric evaluations and need to be confirmed by body composition measurement. Black subjects have a higher FFM than white subjects because of a heavier and denser skeletal mass and denser muscle mass (30). Furthermore, middle-aged African-American men 8

11 and women have lower visceral fat than white subjects for a given waist circumference and BMI (31). In contrast, in a study including 445 white and 242 Asian subjects, Wang et al. showed that Asians were shorter, lighter and had a lower BMI than white European subjects, but had a higher % FM and upper-body subcutaneous fat for a given BMI (32). Another study described also a higher abdominal FM, especially visceral FM in 822 Chinese and South Asian subjects than in Europeans for a given total body fat (33). Regarding smoking, past and current smoking has been associated with increased abdominal fat, evaluated through waist to hip ratio, in large populations of middle-aged and elderly subjects (34, 35). The impact of past and current smoking may be dependent on gender. A Dutch study including over 2200 subjects showed a positive association with abdominal fat in elderly men but not women (36). Lower social educational achievement and lower adult social class are associated with higher BMI in over 2000 Finnish men and women (37). Furthermore, Lahmann et al. described that early socioeconomic status is among the strongest determinants of weight gain and increased waist-to-hip ratio during adult life (38). In women, parity and menopause also affect body composition. For instance, increasing parity is associated with a relative decrease in lower-body fat, but an increase in waist circumference (39) The fat regained after the postpartum period is mostly stored in central vs. peripheral depots. Finally, early menopause is related to an increased intra-abdominal fat (40). 9

12 2. METHODS FOR BODY COMPOSITION MEASUREMENTS Reference methods Generalities The reference methods for measuring FM and FFM in vivo are in vivo neutron activation, underwater weighing, dilution techniques, computed tomography, magnetic resonance imaging, dual-energy x-ray absorptiometry and total body potassium. In view of their complexity and costs, these methods remain presently research tools. In vivo neutron activation relies on administration of radioisotopes to humans. It can measure total body calcium, phosphorus, sodium and chlorine levels. Total body calcium levels allow measurement of skeletal mass. Total body nitrogen allows calculation of total body protein mass (TBP), thus lean tissue, by the relationship: Protein (g) = 6.25 x nitrogen (g) (41). Underwater weighing, or hydrodensitometry, predicts percent body fat from body density (42) by the equation of Siri: Percent body fat = (495 / body density) (43). Body density is derived from body mass and body volume, where body volume is equal to the loss of weight in water, by the equation: Body density = body mass/ [((body mass - body weight underwater) / water temperature correction) - residual lung volume]. The dilution technique allows measurement of total body water by the use of tritiated ( 3 H), deuterated ( 2 H) or oxygen ( 18 O)-labeled water. Orally labeled water is ingested and, once equilibrium is achieved, a sample of urine or blood allows calculation of TBW. FFM is calculated by the relationship TBW/FFM =0.73, which assumes a stable hydration of FFM (44). Finally, computed tomography and magnetic resonance imaging are mainly used to quantify the FM in the visceral, subcutaneous and intramuscular area and the FFM in specific high rate metabolic organs, rather than to measure whole body composition (45). 10

13 Dual-energy x-ray absorptiometry and total body potassium are available in the University Hospital of Geneva and have therefore been used as reference methods in our studies dealing with validation of bedside methods measuring whole body composition. Their functioning is detailed underneath Total body potassium Total body potassium (TBK) is measured with a whole body scintillation 40 K counter (figure 2). The subject, sitting in a tilting chair, is placed in the field of view of a large sodium iodide crystal (203 mm-diameter). The gamma photons emitted by the subject interact with the crystal and send an output signal to the spectrum analyzer. In order to minimize background interference, the subject and the detector are placed in a room surrounded by thick steel walls. 40K is measured by counting the total pulses recorded in the channels of the photo peak of this isotope for 30 minutes and divided by to obtain total body potassium (TBK) in millimoles. At our institution, the accuracy and reproducibility of this method are 5% and 2%, respectively (46). Figure 2: Body scintillation 40K counter 11

14 TBK allows calculation of body cell mass (BCM) and FFM. BCM is the working portion of lean body mass by the following formula: BCM (kg) = x TBK (mmol) (47). This formula relies on the assumptions that the average intracellular potassium content is 3 mmol/g nitrogen and the nitrogen content is 0.04 g/g wet tissue. There is good evidence to support these assumptions in weight stable healthy subjects (48). Calculation of FFM relies on the ratio of TBK/FFM, which equals 68.1 and 64.2 mmol/kg FFM in men and women according to Forbes et al. (49). However, this ratio is actually lower in healthy non obese men (66.3± 5mmol/kg FFM) aged years and it decreases with age (50). Thus, the use of a fixed TBK/FFM ratio is inappropriate. This has led to the development of a new formula to calculate FFM from TBK, derived from multiple regressions and using DXA as reference method (51). FM is obtained by subtraction of FFM from body weight. FFM = (TBK) (height) (weight) (sex), where sex equals 0 in women and 1 in men Dual-energy x-ray absorptiometry Dual-energy X-ray absorptiometry (DXA) relies on the use of an X-ray tube directing two energy levels of photons through the body (figure 3). When passing through the body, these X-ray beams are attenuated due to the absorption and scattering of the photons. The attenuation is measured for every pixel of the body surface by a detector and complex calculations allow determination of bone mineral and soft tissue densities. Soft tissue can further be divided into FM and lean body mass as they have different attenuation characteristics. Its detailed functioning is mentioned in a review (52, appendix 1). DXA 12

15 induces radiations of 2 to 5 µsv, which is low compared to the daily background radiation of 5 to 7 µsv. Figure 3: Dual-energy x-ray absorptiometry Presently, the three most common DXA manufacturers are Hologic Inc, GE-Lunar Inc and Norland Medical Systems. Their devices are based on the same physical principle but differ in their conception. Several studies have mentioned differences in body composition between devices of different brands and devices of the same brand using various softwares. We compared body weight and composition measured by balance beam scale and more recent DXA devices, i.e. the DXA Hologic QDR 4500-A in normal (NPM) and high power mode (HPM) and the Lunar Prodigy in normal and overweight women (53, appendix 2). In obese subjects, the QDR 4500A HPM overestimated scale weight and the QDR 4500A NPM underestimated scale weight in obese women. The Prodigy gave weight about similar to scale. Regarding body composition, both Hologic acquisition modes underestimated FM but overestimated lean body mass compared to Prodigy (table 4). The differences of body weight and FM between these devices increased with high BMI. 13

16 Table 4: Body composition by QDR4500-A and Prodigy (mean ±SD) (adapted from (53)). n Scale weight Lean body Bone mineral Fat mass (kg) mass (kg) content (kg) (kg) 25<BMI<30 kg/m 2 13 Scale QDR4500-A NPM QDR4500-A HPM Prodigy 71.0± ± ± ± ± ± ± ± ± ±0.22* 30<BMI<35 kg/m 2 8 Scale 92.2±11.9 QDR4500-A NPM 90.4± ± ±0.35 QDR4500-A HPM 94.2± ± ±0.34 Prodigy 92.7± ± ± <BMI<40 kg/m 2 8 Scale 89.3±7.5 QDR4500-A NPM 87.1± ± ±0.26 QDR4500-A HPM 91.3± ± ±0.26 Prodigy 89.9± ± ±0.29 BMI>40 kg/m 2 7 Scale 103.8±4.6 QDR4500-A NPM 99.7± ± ±0.12 QDR4500-A HPM 105.9±4.2* 56.4±4.1* 2.13±0.12 Prodigy 104.4±3.9* 49.0± ±0.17* *p<0.05 vs. QDR 4500-A NPM, vs. QDR 4500-A HPM (ANOVA) 26.1± ± ± ± ± ± ± ± ± ± ± ±4.3* The short-term precision of these DXA devices regarding body composition was measured in the same study. The coefficients of variation for lean body mass were 0.94% for the QDR 4500-A NPM, 0.63% for the QDR 4500-A HPM and 1% for the Prodigy. Furthermore, since obese subjects may not fit in the DXA scanning area, we validated the measurement of right half-body DXA by DXA Lunar Prodigy to predict total body composition in subjects with normal BMI (54) and in obese women (55). 14

17 2. 2. Bedside methods Generalities The afore-mentioned reference methods are expensive, require extensive operator training and cannot be performed at the patient s bedside. Thus, in clinical routine, health care providers generally evaluate body composition through BMI, skinfold thickness, circumferences or bioelectrical impedance analysis Skinfold thickness and circumferences Skinfold thickness predicts percent body fat from measurements of subcutaneous fatfolds. Among 50 frequently used equations for FM and FFM prediction, the most common areas for taking skinfold measurements are at the triceps, subscapular, abdominal and iliac crest, thigh, biceps, calf, chest, umbilicus and thorax, on the right side (56). Its main limitation is the interobserver variability, which can be due to the calipers used, the differences in location of the anatomical sites and the technique of grasping the skinfolds. The other limitations are the validation of equations, which were developed in one population but may not be valid in another population, and obesity (57). As measurement of skinfold thickness, circumferences allow prediction of body composition and distribution of FM. They can be measured at the level of the chest, arm, waist, hip, thigh, knee, calf and ankle (56). They have several advantages over skinfold thickness: they can be measured regardless of body size and fatness and they evaluate fat distribution because waist circumference and waist-to-hip ratio are surrogates of abdominal FM (58). However, as for skinfold thickness, their use is limited by the interobserver variability and formulas, which may not be population-specific. Although both methods are relatively cheap, they are not used in clinical routine at the Geneva University Hospital because of their high interobserver variability. 15

18 Bioelectrical impedance analysis Considering the pitfalls of anthropometrics, attention of the nutritional health care providers focuses on bioelectrical impedance analysis (BIA), which is an easy, quick, noninvasive and especially operator-independent method for measuring body composition. The estimation of body composition by BIA is based on the geometrical relationship between impedance (Z), length (L) and volume (V) of an electrical conductor: V = ρl2/z (ρ is the specific resistivity, in ohms). Adapted to the human body, V corresponds to the volume of FFM, and L to the height of the subject. Z is composed of the pure resistance R of the conductor, the FFM, and the reactance Xc produced by the capacitance of cellular membranes, tissue interfaces and nonionic tissues: Z2 = R2 + Xc2 (59). Practically, after cleaning the skin with 70% alcohol, adhesive electrodes are placed on the right hand and foot while the subject is lying on his back as shown on figure 4. A generator applies an alternating electrical current of 50 khz and 0.8 ma to these electrodes. The measured resistance and reactance are used to calculate FFM by formulas validated against reference methods of body composition measurements. The subtraction of FFM to weight allows calculation of FM. Figure 4: Bioelectrical impedance analysis 16

19 The limitations of single-frequency BIA are fluid overload, as ascites or edema, or fluid shifts as in hemo- or peritoneal dialysis, body shape abnormalities and amputations, and extreme BMIs (<16 kg/m 2 or >34 kg/m 2 ). In neuromuscular diseases as Duchenne myopthathy or posttraumatic plegia, BIA needs further validation (59, 60). Its short and long-term of resistance measurements show coefficient of variations of % (61). Numerous BIA formulas have been validated against reference methods and published. However, in healthy adults, different ages, ethnic groups, gender and body shapes may preclude the use of one single formula (59). We have developed and validated against DXA, a single BIA equation using a double cross-validation technique, in healthy white European subjects, aged years and with a BMI of kg/m 2 (61, appendix 3). FFM = (height 2 /resistance) (weight) (reactance) (sex), where sex equals 0 in women and 1 in men. This formula, termed the Geneva (or Kyle) formula, has later on been validated specifically in elderly subjects. In a cross-sectional study, we compared FFM obtained by our non-age-specific equation and three BIA equations developed specifically for elderly subjects, with FFM derived from DXA QDR 4500A (62, appendix 4). Our formula accurately predicted FFM in Swiss subjects between 65 and 94 years, with a BMI of 17 to 34.9 kg/m 2 (figure 5). However, the other age-specific BIA formulas were not valid in our population. 17

20 Figure 5: Bland-Altman plots of FFM measured by DXA and the Geneva BIA formula in healthy elderly subjects (adapted from (62)). The lines show the mean difference ± 2SD between FFM DXA and FFM Geneva. Women Men FFM Geneva - FFM DXA (kg) FFM Geneva - FFM DXA (kg) FFM Geneva - FFM DXA (kg) / 2 FFM Geneva - FFM DXA (kg) / 2 This formula has also been validated against DXA QDR-4500 in patients before and after liver, lung and heart transplantation (63). Noteworthy is that patients with visible edema, or ascites had been excluded as single-frequency BIA is not valid in these situations. The afore-mentioned formulas measure whole body composition. However, several authors have published equations to determine specifically appendicular skeletal muscle mass (ASSM) from anthropometric (64, 65) or BIA parameters (66). ASSM accounts for > 75% of skeletal muscle and is essential for ambulation and physical activity. We have also determined our own formula, validated against DXA in the Geneva population (67, 68). ASSM = (height 2 /resistance) (weight) (age) (reactance) (sex), where sex equals 0 in women and 1 in men 18

21 3. NORMATIVE VALUES OF TOTAL BODY COMPOSITION Swiss normative values The interpretation of body composition values and thus the use of BIA in clinics suppose normative values of body composition determined in healthy subjects. We determined reference values of FM and FFM from BIA measurements performed in 5225 healthy Caucasian subjects aged years, living in the Geneva area (69, appendix 5). FFM was calculated by the Geneva formula, which has been validated against DXA as mentioned before. The results of this cross-sectional study were expressed in percentiles for absolute (kg) and relative (%) values, in age groups of 10 years. They showed that absolute FFM peaked in the age class years in men and years in women and declined thereafter, while weight, FM and % FM increased in both genders between 15 and 98 years (figure 6). Figure 6: Percentiles of FFM in kg (A), FM in kg (B), and FM in % (C) according to age classes (adapted from (69)) A. Men Women FFM (kg) th 75th 50th 25th 10th yrs th 75th 50th 25th 10th yrs 19

22 B. Men Women FM (kg) th 75th 50th 25th 10th th 75th 50th 25th 10th 0 yrs 0 yrs C. Men Women FM (%) th 75th 50th 25th 10th th 75th 50th 25th 10th 10 yrs 10 yrs However, these normative values have their limitations. When expressed in kg, they rely on height. A short subject will obviously have a lower FFM than a tall subject. As a consequence, a short subject may be considered erroneously depleted in FFM. When expressed in %, the normative values depend on body weight. For instance, an obese subject will have a lower % FFM than a normal weight subject of identical absolute FFM. To override these problems, we have pursued by publishing reference values of body composition as FFM and FM indexes, in 6733 healthy Caucasian sedentary and physically active subjects (70, appendix 6). These indexes reflect body composition adjusted for weight 20

23 and are calculated by dividing FFM and FM by height (m) 2. This way of normalization for body height derives from BMI, which is in fact a measure of height-adjusted weight. BMI and FM index increased progressively with aging in men and women. FFM index peaked at age years, remained stable until age years and then decreased in both genders Impact of aging The normative values of body composition in different age classes were described during the preceding chapter. These values relied on BIA measurements with its inherent limitations. However, BIA may not be able to capture fine changes of body composition. Thus, we have gradually tried to better define these changes using the reference methods for body composition measurements available at our institution, i.e. DXA and TBK. In a cross-sectional study including 433 ambulatory Caucasian subjects aged yrs, we have evaluated the differences in FFM, ASMM, BCM and fat between age classes (71). BCM was determined by TBK and the other body composition parameters by DXA. None of the subjects had physical limitations or restrictions in mobility. The decline of lean components was accelerated in men and women after 60 years while FM continued to increase until around 75 years. As a consequence of these findings, we studied especially the changes of body composition in subjects older than 60 years (72, appendix 7). There was a relatively greater decrease of TBK and BCM than of FFM (figure 7) and consequently a decrease of the ratio TBK/FFM with age. This suggested that the composition of FFM was changing after 60 years. Interestingly, physical activity, including household, walking and sport activities, expressed as minutes/d, was positively associated with BCM, ASMM and TBK/FFM but not with FFM or FFM index. 21

24 Figure 7: FFM and BCM decrease according to age classes (mean ± SD) (adapted from (72)) *p < 0.05 vs. the preceding age group (adjusted for height and weight) (ANOVA) Men Women FFM (kg) * FFM (kg) BCM (kg) * * BCM (kg) * * yrs yrs > 80 yrs yrs yrs > 80 yrs As a consequence of this dissimilar decrease, the prevalence of sarcopenia was changing. Indeed, sarcopenia can be defined either by -2SD of the BCM index (BCM/height (m) 2 ) or -2SD of the Relative Skeletal Muscle Index (RSM=ASMM/height (m) 2 ) of a reference population. In this study, 11 % of men and 11% of women were classified as sarcopenic using the RSM index but 45% of men and 30 % of women using the BCM index. Thus, the prevalence of sarcopenia was much higher when using the BCM index showing that the definition of sarcopenia had to be refined for elderly subjects Impact of physical activity Physical activity influences weight and body composition. In our first cross-sectional study on this topic, 3853 healthy subjects aged 15 to 64 years underwent BIA measurements (73). Subjects were classified as sedentary or physically active on the basis of their weekly physical activity. Physically active subjects were those performing over 3 hours/week of an endurance-type physical activity classified with an intensity code of 4 metabolic equivalents 22

25 (MET) or more according to the Minnesota Leisure Time Activities Questionnaire (74). The other subjects were classified as sedentary. Weight and FM increased with aging in both sedentary and physically active subjects. Physically active subjects had no increased FFM compared to their sedentary counterparts, but less weight and FM in all age categories. When extending the study to subjects aged 18 to 98 years and adjusting body composition values for height, we confirmed that, as expected, FM index was higher in sedentary than physically active subjects (70). The differences in FM index between both groups increased progressively with age. This study demonstrated that, in physically active subjects, FFM index was stable until 74 years, and thereafter tended to decrease in men and increase in women. However, the subjects aged over 75 years were few and thus the results needed confirmation. In order to get an insight in the impact of physical activity changes on body composition, we followed the body composition of 213 healthy subjects aged over 65 years in a longitudinal study over 3 years. Despite a stable body weight, fat-free soft tissue measured by DXA (FFM bone mineral mass), ASSM and TBK / fat-free soft tissue decreased over three years. In contrast, FM, especially in the abdominal area, increased. Multiple regression analysis showed that body composition changes were related mainly to body weight changes but not to sex and age. At inclusion, subjects were classified as physically active if they spent over 10% of their total energy expenditure in activities over 4 METs, or as sedentary. With this kind of classification occurring at one time-point, we found no effect of physical activity on changes of fat-free soft tissue, ASSM, TBK/ fat-free soft tissue, FM and truncal FM (75). The afore-mentioned follow-up over 3 years was continued over 9 years, in all subjects who could be contacted and were healthy enough to undergo the measurements of body composition (n=112) (76, appendix 8). Several health scores (Charlson score, Mini Mental State Examination, Geriatric Depression Scale, Barthel index) were performed at follow-up 23

26 and confirmed the good health of the subjects who continued the study. This recently published paper confirmed that the loss of FFM did not occur progressively with aging but arose around 70 years. A validated physical activity questionnaire was performed at inclusion and at the 9-year follow-up visit. Subjects were classified into one of two physical activity categories: those who increased or those who decreased energy expenditure through physical activity. During these years, the men, but not the women, who increased their energy expenditure through physical activity limited the loss of FFM (figure 8). Figure 8: Changes of body composition according to changes in physical activity expenditure in men (A) and women (B) (adapted from (76)). = difference between 2008 and Subjects increasing and subjects decreasing physical energy expenditure. *p<0.05 between both groups (unpaired t-test). A (kg) * * * * * Weight FM FFM ASMM BCM 24

27 B (kg) Weight FM FFM ASMM BCM Impact of obesity As mentioned in the chapter on bioelectrical impedance, obesity may affect the validity of BIA formulae. To confirm this often read statement, we compared FFM obtained by the Geneva BIA formula and two BIA equations developed specifically in obese subjects with FFM derived from DXA Prodigy, in 26 obese women with a BMI of 36.7±4.8 kg/m 2 (77). Since DXA Hologic did not measure accurately weight compared to scale in our former study comparing DXA devices in obese subjects, we used DXA Prodigy as a reference method. The obese-specific formulas of Segal et al. (78) and Gray et al. (79) compared better to DXA Prodigy than the Geneva BIA formula, validated against DXA Hologic. The Geneva formula overestimated FFM in obese subjects, as shown in a Bland-Altman analysis (figure 9). Noteworthy is that our BIA formula was developed against the DXA Hologic and not the Prodigy. Thus, the overestimation of the Geneva BIA formula relied on the reference DXA device. 25

28 Figure 9: Bland-Altman analysis of FFM measured by DXA Prodigy and the BIA formulas of Geneva (A), Segal et al. (B) and Gray et al. (C) in obese subjects. The lines show the mean differences ± 2SD between FFM DXA and FFM BIA measured by the formulas of Geneva, Segal et al. and Gray et al.. A. B. FFM DXA Prodigy FFM BIA Geneva FFM DXA Prodigy FFM BIA Segal FFM DXA Prodigy + FFM BIA Geneva /2 FFM DXA Prodigy + FFM BIA Segal /2 C. 8 FFM DXA Prodigy FFM BIA Gray FFM DXA Prodigy + FFM BIA Gray /2 We further evaluated whether % FM, measured by DXA Prodigy, or BMI determined inaccuracy of BIA-derived FFM (80). BMI affected the accuracy of FFM calculated by the 26

29 obese-specific formula of Segal et al., and %FM the accuracy of the formula of Gray et al.. However, neither BMI nor %FM affected FFM calculated by the Geneva BIA formula, suggesting the possibility of applying a correction factor in obese women, independently of BMI or % FM Impact of environment One year after the publication of the Geneva formula, Chumlea et al. published reference values of body composition for North American subjects, in age groups of 10 years (81). These values were obtained through BIA measurements in subjects examined in the NHANES III (National Health and Nutrition Examination Survey), and the subsequent use of a newly developed BIA formula, the NHANES equation. Included subjects were non- Hispanic whites, non-hispanic blacks and Mexican-Americans. In order to evaluate the adequacy of this formula compared to ours, we compared the FFM measured by the NHANES and the Geneva formula, in healthy Swiss adults, aged 20 to 79 years (82). Mean FFM measured by the Geneva formula was not different from that measured by the NHANES formula in men, but significantly lower in women (-1.5 kg). In the same study, we pursued by evaluating the differences between North American non-hispanic white subjects, whose body composition was calculated with the NHANES formula, and Swiss subjects, whose body composition was calculated with the Geneva formula. We came to the conclusion that in almost all age class, American subjects yield higher weight, FFM, FM and % FM. Besides geographical environment, body composition is also influenced by modifications of environment related to time. Between 1993 and 2003, the prevalence of subjects with overweight, high FFM index and high FM index increased in Switzerland (83). Leisure-time physical activity was negatively associated with BMI, FFM index and FM index. 27

30 4. CLINICAL USE OF BODY COMPOSITION Nutritional risk Nutritional state is evaluated through clinical history, medical examination, biological markers, anthropometrics and body composition. A decrease of FFM is a hallmark of protein energy malnutrition and has been related with many complications as described in chapter However, the cut-offs of anthropometrics and FFM under which the patient is at risk for complications is unclear and may be different depending on the primary disease or the considered end-point. A review published a few years ago summarized the studies dealing with this subject (84). In our institution, we have evaluated especially the link between nutritional parameters and length of hospital stay. In 995 consecutive patients admitted to the hospital admission center and subsequently hospitalized, nutritional state was evaluated by BMI, serum albumin, Subjective Global Assessment and body composition measured by 50-kHz bioelectrical impedance (85). Subjective Global Assessment determines nutritional state based on features as weight change, dietary intake change, gastrointestinal symptoms, functional capacity, disease and its relation to nutritional requirements, and physical examination. It classifies subjects as well-nourished (score A), moderately malnourished (score B) and severy malnourished (score C) (86). Body composition was compared to that of age- and heightmatched healthy volunteers. The prevalence of malnutrition depended on the parameter used to define malnutrition. In patients younger and older than 60 years, we found a BMI 20 kg/m 2 in 20 and 14%, respectively, a serum albumin 34.9 g/l in 7 and 24%, respectively, and a Subjective Global Assessment (SGA) showing moderate or severe depletion in 50 and 79%, respectively. The FFM decreased with worsening of Subjective Global assessment and was lower in patients older than 60 years compared to their younger counterparts (table 5). This study did not 28

31 determine a clear cut-off of health risk related to FFM loss but showed that body composition measurements helped identifying moderately or severely depleted patients at hospital admission. Furthermore, it highlights the high prevalence of malnutrition at hospital admission, regardless of the used parameter, suggesting that malnutrition may be a risk for hospital admission. Table 5: Body composition parameters according to SGA (adapted from (85)) Volunteers Well-nourished Moderately depleted Severely depleted Men 60 yrs (n Weight FFM 74.8± ± ± ± ±10.1* 55.0±6.2* 58.8±9.2* 48.8±6.2* Men > 60 yrs Weight FFM 74.2± ± ± ± ± ± ±12.4* 46.7±7.4* Women 60 yrs Weight FFM 59.2± ± ± ± ±8.9* 40.5±4.1* 51.2±9.5* 36.0±4.6* Women > 60 yrs Weight FFM 62.5± ± ± ± ±10.2* 37.3±5.3* 51.6±10.1* 33.1±4.8* *p vs. well-nourished (ANOVA). No statistical significance of severely depleted vs. moderately depleted nor or well-nourished vs. volunteers. In a following article, my colleagues classified the subjects into categories of low, normal and high FFM indexes. This classification relied on former published reference values of FFM indexes for the BMI categories defined by the World Health Organization (87). The FFM index of subjects with a BMI < 20 kg/m 2 was considered low and corresponded to a value < 17.4 kg/m 2 in men and < 15.0 kg/m 2 in women. Subjects with a low FFM index had a higher length of stay than patients with a normal or high FFM index (14). FFM index was a 29

32 more sensitive determinant of length of stay than weight loss > 10% or BMI < 20 kg/m 2. These results suggested that that FFM index should be used to evaluate nutritional risk. A subsequent study included 1707 patients from Geneva and Berlin who underwent body composition measurements within 24h of admission. The study showed that a high FM index, and not only a low FFM index, was associated with length of stay (88) Nutritional follow-up Body composition measurements are not only useful to assess nutritional risk. This chapter reports three published clinical situations where longitudinal body composition measurements helped guiding nutritional treatment and understanding metabolism in patients with complex diseases. In the first situation, body composition measurements helped guiding nutritional support in a woman with a short bowel syndrome (89, appendix 9). She had a history of Roux-en-Y gastric bypass complicated by a small bowel volvulus and was left with 25 cm of ileum. Weight and body composition measurements allowed adaptation of parenteral nutrition. They also differentiated variations of FM vs. FFM, which is important as loss of FFM has been associated to malnutrition-related complications. The second case concerned a 13-year old boy, who underwent a living-related small bowel transplantation from his monozygotic twin brother (90, appendix 10). He was regularly assessed for nutritional state in the pre- and post-operative period through BIA and DXA measurements. He had no immunosuppressive therapy after the operation. The graft recipient caught up with the height and FFM of his brother within two years of small bowel transplantation, but his FM and femoral bone densities were lower than in the donor after four years (figure 10). Thus, we could conclude that small bowel transplantation improved nutritional state and growth of the graft recipient although body composition had not totally 30

33 normalized after four years. Figure 10: Evolution of height (A), weight (B) and body composition (C and D) after bowel transplantation (adapted from (90)). Day O: day of bowel transplantation. Transplant receiver. Transplant donor. A. B Height (cm) Weight (kg) days days C. D Fat-free mass (kg) Fat mass (kg) days days Finally, variations of body composition were studied in patients before and after lung transplantation. Before lung transplantation, they had lower FM and FFM than healthy age and height-matched volunteers. After transplantation they slowly increased FFM and FM. However, at 24 months, their FFM was still lower than that of volunteers. This study showed that the nutritional state of lung transplanted patients improved after transplantation but had still not returned to normal level after 24 months (91). 31

34 5. SUMMARY AND CONCLUSION Measurement of body composition is an important part of nutritional assessment. The low FFM associated with malnutrition has been associated with numerous infectious and noninfectious complications, increasing length of stay, morbidity and mortality. DXA and TBK are reference method for determination of FFM and BCM, but, as these methods are expensive and require extensive technique of the operator, we have focused especially on BIA, an easy, quick, safe and reliable bedside method to measure body composition. BIA formulas to routinely assess FFM and appendicular skeletal muscle mass have been developed. Normative values of total body composition have been established, according to age and gender. Longitudinal and cross-sectional studies allowed an insight on the impact of physical activity and environment on body composition. With regard to clinics, we have studied the impact of body composition, determined by BIA, on length of hospital stay and shown that a low FFM and FM index were associated with an increased length of stay. This demonstrates that prevention of FFM loss, whether through nutritional support or drugs, may improve clinical outcome and decrease hospital costs. Finally, we have shown two case reports where sequential FFM measurements guided nutritional and medical therapy. Future studies should focus on the relationship between body composition and outcome in various types of patients. They should also try to determine the characteristics of nutritional support (amount of calories, type of macronutrient intakes, timing of nutritional support), potentially associated with anabolic drugs or physical activity, necessary to limit FFM loss, in order to improve clinical outcome. 32

35 ACKNOWLEDGEMENTS This work could not have been done without the precious support of a whole team. My deepest thanks go to all the people with whom I had a chance to work, whether in clinics or in research and especially my colleagues from the nutrition unit and the dieticians, at the University Hospital of Geneva. Among these, I am especially grateful to: - Prof. Claude Pichard, who put a lot of himself in my training, whether in clinical nutrition or in research. He guided me in order to get the best out of my training, whether abroad or in Geneva, was always open for teaching session and scientific advice, and stimulated my creativity in research with a lot of open-mindness. - Laurie Karsegard, my dearest confident, who had always an open ear when I needed it. She encouraged and motivated me throughout my training. She was and still is a key person for me, with whom I very much appreciate to work. - Ursula Kyle, who is a model of creativity and perseverance for me. I loved the critical interactions with her when sharing ideas and I am missing them since she moved to the USA. Without family support, it would have been impossible to ally professional life and motherhood. I would therefore like to thank all grand-parents, and especially my mother who encouraged me in pursuing my career and stepped in whenever I needed her for the kids, and my father, who always supported my professional choices. and of course, my grateful thoughts go to Christophe, my husband, and to Ludovic and Céliane, my kids, whose smiles and hugs always remind me the priorities of life. 33

36 REFERENCES 1. Wang ZM, Pierson RN, Jr., Heymsfield SB. The five-level model: a new approach to organizing body-composition research. Am J Clin Nutr 1992;56: Bemben MG, Massey BH, Bemben DA, Boileau RA, Misner JE. Age-related variability in body composition methods for assessment of percent fat and fat-free mass in men aged years. Age Ageing 1998;27: Melchior JC, Thuillier F. Méthodes d'évaluation de l'état nutritionnel. In: Cano NJ, Barnoud D, Schneider S, Vasson M-P, Hasselmann M, Leverve X, eds. Traité de nutrition artificielle de l'adults. Paris: Springer, Genton L, van Gemert W, Pichard C, Soeters P. Physiological functions should be considered as true end points of nutritional intervention studies. Proc Nutr Soc 2005;64: Morabia A, Ross A, Curtin F, Pichard C, Slosman DO. Relation of BMI to a dualenergy X-ray absorptiometry measure of fatness. Br J Nutr 1999;82: Kyle UG, Janssens JP, Rochat T, Raguso CA, Pichard C. Body composition in patients with chronic hypercapnic respiratory failure. Respir Med 2006;100: Evans WJ, Morley JE, Argiles J, et al. Cachexia: a new definition. Clin Nutr 2008;27: Thomas DR. Loss of skeletal muscle mass in aging: examining the relationship of starvation, sarcopenia and cachexia. Clin Nutr 2007;26: Evans WJ. Skeletal muscle loss: cachexia, sarcopenia, and inactivity. Am J Clin Nutr 2010;91:1123S-1127S. 10. Janssen I, Heymsfield SB, Ross R. Low relative skeletal muscle mass (sarcopenia) in older persons is associated with functional impairment and physical disability. J Am Geriatr Soc 2002;50:

37 11. Giron R, Matesanz C, Garcia-Rio F, et al. Nutritional state during COPD exacerbation: clinical and prognostic implications. Ann Nutr Metab 2009;54: Cano NJ, Roth H, Court-Ortune I, et al. Nutritional depletion in patients on long-term oxygen therapy and/or home mechanical ventilation. Eur Respir J 2002;20: Chandra RK. Nutrition, immunity and infection: from basic knowledge of dietary manipulation of immune responses to practical application of ameliorating suffering and improving survival. Proc Natl Acad Sci U S A 1996;93: Pichard C, Kyle UG, Morabia A, Perrier A, Vermeulen B, Unger P. Nutritional assessment: lean body mass depletion at hospital admission is associated with an increased length of stay. Am J Clin Nutr 2004;79: Hitzl AP, Jorres RA, Heinemann F, Pfeifer M, Budweiser S. Nutritional status in patients with chronic respiratory failure receiving home mechanical ventilation: impact on survival. Clin Nutr 2010;29: van Wetering CR, Hoogendoorn M, Broekhuizen R, et al. Efficacy and costs of nutritional rehabilitation in muscle-wasted patients with chronic obstructive pulmonary disease in a community-based setting: a prespecified subgroup analysis of the INTERCOM trial. J Am Med Dir Assoc 2010;11: Guest JF, Panca M, Baeyens J-P, et al. Health economic impact of managin patients following a community-based diagnosis of malnutrition in the UK. Clinical Nutrition 2011;In press. 18. Organization WH. Obesity: preventing and managing the global epidemic: report of a WHO consultation. Geneva: World Health Organization, Houston DK, Stevens J, Cai J. Abdominal fat distribution and functional limitations and disability in a biracial cohort: the Atherosclerosis Risk in Communities Study. Int J Obes (Lond) 2005;29:

38 20. Miller WC, Kocejy DM, Hamilton EJ. A meta-analysis of the past 25 years of weight loss research using diet, exercise or diet plus exercise intervention. Int J Obes Relat Metab Disord 1997;21: Garrow JS, Summerbell CD. Meta-analysis: effect of exercise, with or without dieting, on the body composition of overweight subjects. Eur J Clin Nutr 1995;49: Ballor DL, Poehlman ET. Exercise-training enhances fat-free mass preservation during diet-induced weight loss: a meta-analytical finding. Int J Obes Relat Metab Disord 1994;18: Chaston TB, Dixon JB. Factors associated with percent change in visceral versus subcutaneous abdominal fat during weight loss: findings from a systematic review. Int J Obes 2008;32: Sims EA, Goldman RF, Gluck CM, Horton ES, Kelleher PC, Rowe DW. Experimental obesity in man. Trans Assoc Am Physicians 1968;81: Forbes GB. Body fat content influences the body composition response to nutrition and exercise. Ann N Y Acad Sci 2000;904: Rozenek R, Ward P, Long S, Garhammer J. Effects of high-calorie supplements on body composition and muscular strength following resistance training. J Sports Med Phys Fitness 2002;42: Fiatarone MA, O'Neill EF, Ryan ND, et al. Exercise training and nutritional supplementation for physical frailty in very elderly people. N Engl J Med 1994;330: Meredith CN, Frontera WR, O'Reilly KP, Evans WJ. Body composition in elderly men: effect of dietary modification during strength training. J Am Geriatr Soc 1992;40: Genton L, Melzer K, Pichard C. Energy and macronutrient requirements for physical 36

39 fitness in exercising subjects. Clin Nutr 2010;29: Conway JM, Yanovski SZ, Avila NA, Hubbard VS. Visceral adipose tissue differences in black and white women. Am J Clin Nutr 1995;61: Carroll JF, Chiapa AL, Rodriquez M, et al. Visceral fat, waist circumference, and BMI: impact of race/ethnicity. Obesity (Silver Spring) 2008;16: Wang J, Thornton JC, Russell M, Burastero S, Heymsfield S, Pierson RN, Jr. Asians have lower body mass index (BMI) but higher percent body fat than do whites: comparisons of anthropometric measurements. Am J Clin Nutr 1994;60: Lear SA, Humphries KH, Kohli S, Chockalingam A, Frohlich JJ, Birmingham CL. Visceral adipose tissue accumulation differs according to ethnic background: results of the Multicultural Community Health Assessment Trial (M-CHAT). Am J Clin Nutr 2007;86: Barrett-Connor E, Khaw KT. Cigarette smoking and increased central adiposity. Ann Intern Med 1989;111: Troisi RJ, Heinold JW, Vokonas PS, Weiss ST. Cigarette smoking, dietary intake, and physical activity: effects on body fat distribution--the Normative Aging Study. Am J Clin Nutr 1991;53: Visser M, Launer LJ, Deurenberg P, Deeg DJ. Past and current smoking in relation to body fat distribution in older men and women. J Gerontol A Biol Sci Med Sci 1999;54:M Salonen MK, Kajantie E, Osmond C, et al. Role of socioeconomic indicators on development of obesity from a life course perspective. J Environ Public Health 2009;2009: Lahmann PH, Lissner L, Gullberg B, Berglund G. Sociodemographic factors associated with long-term weight gain, current body fatness and central adiposity in 37

40 Swedish women. Int J Obes Relat Metab Disord 2000;24: Lassek WD, Gaulin SJ. Changes in body fat distribution in relation to parity in American women: a covert form of maternal depletion. Am J Phys Anthropol 2006;131: Toth MJ, Tchernof A, Sites CK, Poehlman ET. Menopause-related changes in body fat distribution. Ann N Y Acad Sci 2000;904: Cohn SH, Vartsky D, Yasumura S, et al. Compartmental body composition based on total-body nitrogen, potassium, and calcium. Am J Physiol 1980;239:E Akers R, Buskirk ER. An underwater weighing system utilizing "force cube" transducers. J Appl Physiol 1969;26: Siri WE. The gross composition of the body. Adv Biol Med Phys 1956;4: Ritz P. Body water spaces and cellular hydration during healthy aging. Ann N Y Acad Sci 2000;904: Lee SY, Gallagher D. Assessment methods in human body composition. Curr Opin Clin Nutr Metab Care 2008;11: Wenger P, Soucas L. Anthropogammamètre (whole body counter) de Genève. Etalonnage pour le potassium naturel et variation de la teneur en potassium naturel en fonction du poids du sujet. Helv Chim Acta 1964;47: Moore FD, Oleson KH, McMurrey JD, Parker HV, Ball MR, Boyden CM. The body cell mass and its supporting environment. Philadelphia: W.B. Saunders, Cohn SH, Vaswani AN, Yasumura S, Yuen K, Ellis KJ. Assessment of cellular mass and lean body mass by noninvasive nuclear techniques. J Lab Clin Med 1985;105: Forbes GB, Hursh JB. Age and Sex Trends in Lean Body Mass Calculated from K40 Measurements: with a Note on the Theoretical Basis for the Procedure. Ann N Y Acad 38

41 Sci 1963;110: Kyle UG, Genton L, Slosman DO, Pichard C. Fixed total body potassium (TBK)/fatfree mass (FFM) ratio does not permit prediction of fat-free mass. Clin Nutr 2000;19:S Kyle UG, Genton L, Slosman DO, Pichard C. Validation of a total body potassium (TBK) equation to predict FFM. FASEB 2001;15: Genton L, Hans D, Kyle UG, Pichard C. Dual-energy X-ray absorptiometry and body composition: differences between devices and comparison with reference methods. Nutrition 2002;18: Genton L, Karsegard VL, Zawadynski S, et al. Comparison of body weight and composition measured by two different dual energy X-ray absorptiometry devices and three acquisition modes in obese women. Clin Nutr 2006;25: Hans D, Genton L, Conicella G, Karsegard L, Pichard C, Slosman DO. Half-body dual x-ray absorptiometry (DXA) predicts whole body composition (WBC): a potential method to measure obese patients. Clin Nutr 2001;20:S Genton L, Hans D, Kyle UG, Karsegard L, Slosman DO, Pichard C. Half-body dual x- ray absorptiometry (DXA) predicts whole body composition (WBC) in obese patients. Clin Nutr 2003;21:S Wang J, Thornton JC, Kolesnik S, Pierson RN, Jr. Anthropometry in body composition. Ann N Y Acad Sci 2000;904: Mattsson S, Thomas BJ. Development of methods for body composition studies. Phys Med Biol 2006;51:R203-R Ness-Abramof R, Apovian CM. Waist circumference measurement in clinical practice. Nutr Clin Pract 2008;23: Kyle UG, Bosaeus I, De Lorenzo AD, et al. Bioelectrical impedance analysis--part I: 39

42 review of principles and methods. Clin Nutr 2004;23: Kyle UG, Bosaeus I, De Lorenzo AD, et al. Bioelectrical impedance analysis-part II: utilization in clinical practice. Clin Nutr 2004;23: Kyle UG, Genton L, Karsegard L, Slosman DO, Pichard C. Single prediction equation for bioelectrical impedance analysis in adults aged years. Nutrition 2001;17: Genton L, Karsegard VL, Kyle UG, Hans DB, Michel JP, Pichard C. Comparison of four bioelectrical impedance analysis formulas in healthy elderly subjects. Gerontology 2001;47: Kyle UG, Genton L, Mentha G, Nicod L, Slosman DO, Pichard C. Reliable bioelectrical impedance analysis estimate of fat-free mass in liver, lung, and heart transplant patients. JPEN J Parenter Enteral Nutr 2001;25: Baumgartner RN, Koehler KM, Gallagher D, et al. Epidemiology of sarcopenia among the elderly in New Mexico. Am J Epidemiol 1998;147: Lee RC, Wang Z, Heo M, Ross R, Janssen I, Heymsfield SB. Total-body skeletal muscle mass: development and cross-validation of anthropometric prediction models. Am J Clin Nutr 2000;72: Janssen I, Heymsfield SB, Baumgartner RN, Ross R. Estimation of skeletal muscle mass by bioelectrical impedance analysis. J Appl Physiol 2000;89: Kyle UG, Genton L, Hans D, Pichard C. Validation of a bioelectrical impedance analysis equation to predict appendicular skeletal muscle mass (ASMM). Clin Nutr 2003;22: Genton L, Kyle UG, Hans D, Pichard C. Validation of a bioelectrical impedance equation to predict appendicular skeletal muscle mass (ASSM). Acta Diabetologica 2002;39:

43 69. Kyle UG, Genton L, Slosman DO, Pichard C. Fat-free and fat mass percentiles in 5225 healthy subjects aged 15 to 98 years. Nutrition 2001;17: Kyle UG, Genton L, Gremion G, Slosman DO, Pichard C. Aging, physical activity and height-normalized body composition parameters. Clin Nutr 2004;23: Kyle UG, Genton L, Hans D, Karsegard L, Slosman DO, Pichard C. Age-related differences in fat-free mass, skeletal muscle, body cell mass and fat mass between 18 and 94 years. Eur J Clin Nutr 2001;55: Kyle UG, Genton L, Hans D, et al. Total body mass, fat mass, fat-free mass, and skeletal muscle in older people: cross-sectional differences in 60-year-old persons. J Am Geriatr Soc 2001;49: Kyle UG, Gremion G, Genton L, Slosman DO, Golay A, Pichard C. Physical activity and fat-free and fat mass by bioelectrical impedance in 3853 adults. Med Sci Sports Exerc 2001;33: Taylor HL, Jacobs DR, Schucker B, Knudsen J, Leon AS, Debacker G. A questionnaire for the assessment of leisure time physical activities. J Chron Dis 1978;31: Raguso CA, Kyle U, Kossovsky MP, et al. A 3-year longitudinal study on body composition changes in the elderly: role of physical exercise. Clin Nutr 2006;25: Genton L, Karsegard VL, Chevalley T, Kossovsky MP, Darmon P, Pichard C. Body composition changes over 9 years in healthy elderly subjects and impact of physical activity. Clin Nutr Genton L, Karsegard L, Hans D, Pichard C. Comparison of fat-free mass (FFM) measured by three bioimpedance analysis (BIA) formula with dual-energy x-ray absorptiometry (DXA) in obese subjects. Clin Nutr 2010;5:S16. 41

44 78. Segal KR, Van Loan M, Fitzgerald PI, Hodgdon JA, Van Itallie TB. Lean body mass estimation by bioelectrical impedance analysis: a four-site cross-validation study. Am J Clin Nutr 1988;47: Gray DS, Bray GA, Gemayel N, Kaplan K. Effect of obesity on bioelectrical impedance. Am J Clin Nutr 1989;50: Genton L, Karsegard L, Hans D, Pichard C. Severity of obesity affects accuracy of fatfree mass (FFM) measured by three bioimpedance analysis (BIA) formula. Clin Nutr 2010;5: Chumlea WC, Guo SS, Kuczmarski RJ, et al. Body composition estimates from NHANES III bioelectrical impedance data. Int J Obes Relat Metab Disord 2002;26: Kyle UG, Genton L, Lukaski HC, et al. Comparison of fat-free mass and body fat in Swiss and American adults. Nutrition 2005;21: Kyle UG, Kossovsky MP, Genton L, Pichard C. Overweight and obesity in a Swiss city: 10-year trends. Public Health Nutr 2007;10: Genton L, van Gemert WG, Dejong CH, Cox-Reijven PL, Soeters PB. When does malnutrition become a risk? Nestle Nutr Workshop Ser Clin Perform Programme 2005;10:73-84; discussion Kyle UG, Unger P, Mensi N, Genton L, Pichard C. Nutrition status in patients younger and older than 60 y at hospital admission: a controlled population study in 995 subjects. Nutrition 2002;18: Detsky AS, McLaughlin JR, Baker JP, et al. What is subjective global assessment of nutritional status? JPEN J Parenter Enteral Nutr 1987;11: Kyle UG, Schutz Y, Dupertuis YM, Pichard C. Body composition interpretation. Contributions of the fat-free mass index and the body fat mass index. Nutrition 42

45 2003;19: Kyle UG, Pirlich M, Lochs H, Schuetz T, Pichard C. Increased length of hospital stay in underweight and overweight patients at hospital admission: a controlled population study. Clin Nutr 2005;24: Genton L, Nardo P, Huber O, Pichard C. Parenteral nutrition independence in a patient left with 25 cm of ileum and jejunum: a case report. Obes Surg 2010;20: Genton L, Raguso CA, Berney T, Hans DB, Morel P, Pichard C. Four year nutritional follow up after living related small bowel transplantation between monozygotic twins. Gut 2003;52: Kyle UG, Raguso CA, Nicod L, Genton L, Pichard C. Longitudinal changes in fat-free (FFM) and fat mass (FM) following lung transplantation. Clin Nutr 2002;21:S84. 43

46 APPENDIX 1

47 REVIEW ARTICLE Dual-Energy X-ray Absorptiometry and Body Composition: Differences Between Devices and Comparison With Reference Methods Laurence Genton, MD, Didier Hans, PhD, Ursula G. Kyle, MS, RD, Claude Pichard, MD, PhD Geneva University Hospital, Geneva, Switzerland INTRODUCTION Body composition measurements provide essential information for assessing and monitoring nutrition state. 1 Some researchers use underwater weighing or potassium counting as reference methods for total body composition. However, these methods assume a constant density and potassium content of lean body mass (LBM), which may not be true, and measure only two compartments, fat (FM) and LBM. 2 A multicompartment approach combining different technologies that measure bone, mineral, muscle and water is preferable and currently considered the gold standard. 3 Nevertheless, its high costs, long duration, and the potential intolerance of patients limit its use in clinical routine. Thus, other methods, especially dual-energy x-ray absorptiometry (DXA), have been investigated. Although the original purpose of DXA was to determine bone mineral density, recent devices measure total and regional body composition of three compartments, fat and lean soft tissues and bone mineral. 4 This editorial focuses on the advantages and limitations of DXA and the differences in total and regional body compositions. In this article, the combination of lean soft tissue and bone mineral is referred to as LBM. BASIC PRINCIPLES The DXA instruments from different manufacturers are unique in implementation but based on the same theoretical principles. DXA devices are composed principally of a generator emitting x rays of two energies, a scanning table, a detector, and a computer system. The fundamental physical principle behind DXA is the measurement of the transmission through the body of x-rays with high- and low-photon energies. When the x-rays of initial photon intensity (I 0, in kev) pass through the body, they are attenuated by photoelectric absorption and Compton scattering, and the intensity transmitted to the detector (I, in kev) is reduced. 5 The attenuation (I/I 0 ) depends on the mass attenuation coefficient ( ) and the areal density of the tissue (M, in g/cm 2 ) 6 : I/I 0 exp(- M) In pixels containing bone (40% of total pixels in the average subject), DXA determines areal densities of bone (M bone ) and soft tissue mass (M soft tissue ) by using two different energies. The following two equations, where the primed variables indicate the low-energy beam and the non-primed the high-energy beam, can be solved, assuming that of bone and soft tissue are known and constant at both energies: This study was supported by Foundation Nutrition 2000 Plus. Correspondence to: Claude Pichard, MD, PhD, Head of Clinical Nutrition and Diet Therapy, Geneva University Hospital, 1211 Geneva 14, Switzerland. pichard@cmu.unige.ch (I/I 0 ) exp ( bone M bone soft tissue M soft tissue ) (I/I 0 ) exp ( bone M bone soft tissue M soft tissue ) Bone mineral density is obtained by averaging M bone of all pixels, and bone mineral content is obtained by multiplying bone mineral density by the projected area. Because there are only two equations to calculate two unknowns (bone and soft tissue mass), no direct calculation of body composition is possible for those pixels that include bone. Percentages of fat and lean soft tissue are deduced from neighboring bone-free pixels, as explained below. 3 In bone-free pixels, DXA directly measures percentages of fat and lean soft tissue (M bone 0). Because the composition of these compartments differs between individuals, no preset values of are available and the aforementioned equations cannot be solved. Thus, DXA compares the ratio of photon attenuation at the lowand high-energy beams (R st ) to experimental R values of fat and lean soft tissue, established by calibration with phantoms equivalent to 100% fat and 100% lean soft tissue. 7 A high R st corresponds to a high proportion of lean soft tissue. 5 An extensive review on the physical and mathematical principles of DXA is provided by Pietrobelli et al. 6 ADVANTAGES AND LIMITATIONS DXA measurements are quick (5 to 10 min), non-invasive, precise, and operator independent. They induce a radiation of 2 to 5 Sv, 8 which is low compared with the daily background radiation of 5 to 7 Sv. 9 Instead of the two compartments measured by underwater weighing and potassium counting, DXA measures three compartments, allowing a better estimation of body composition. However the main advantage of DXA remains the capacity for regional analysis, a fascinating research topic. Risks of cardiovascular disease seem to increase with high abdominal, especially visceral FM, 10 whereas ostoeoarthritis of the knee has been associated with a reduced ratio of leg LBM to body weight. 11 The limitations of DXA come partly from the assumptions this method is based on. First, DXA assumes the same amount of fat over bone as over neighboring bone-free tissue. 9 Indeed, soft tissue determination may not be as accurate in the arms, legs, and thorax as in bone-free regions of interests because only few bone-free pixels are available for direct fat and lean mass calculations. 9 Second, increasing tissue thickness decreases R st by a phenomenon called beam hardening: photons of low energy are removed from the radiation beam, thereby shifting the spectral distribution to higher effective photon energies. DXA assumes that designed calibration phantoms correct for beam hardening. 5 However, this may not be true in very obese subjects, where an underestimation of FM has been observed. 8,12 Third, DXA assumes a constant hydration and electrolyte content of lean soft tissue. 3,13 Although a range of LBM hydration from 68.2% to 78.2% does not significantly alter total percentage of fat, 5 a severe overhydration, such as ascites or edema, may affect R st and resulting percentage of fat. 13 Nutrition 18:66 70, /02/$22.00 Elsevier Science Inc., Printed in the United States. All rights reserved. PII S (01)

48 Nutrition Volume 18, Number 1, 2002 Dual-Energy X-ray Absorptiometry 67 From a practical point of view, in obese patients, the scanner table may be too small to have a complete acquisition of the whole body, leading to inaccurate results. For example, Hendel et al. found an underestimation of body weight by more than 3 kg compared with a high-precision scale in patients heavier than 95 kg. 14 To handle this problem, a half-body scan may be performed 15,16 or the position of the subject changed by putting the arms up and behind the head. 17 COMPARISON OF DXA DEVICES AND THE RESULTING BODY COMPOSITION At present, three manufacturers provide DXA devices measuring body composition: Hologic Inc. (Waltham, MA, USA), Lunar Radiation Corp. (Madison, WI, USA), and Norland Medical Systems (Fort Atkinson, WI, USA). Although based on the same physical principles, differences exist in the generation of the highand low-energy x-ray beams (k-edge filtration versus switching kvp systems), the x-ray detectors, the imaging geometry (pencil versus fan beam versus narrow beam), and the calibration methodology. Additional differences exist in the algorithms used for selective tissue imaging, edge detection, region-of-interest definition, and system calibration. Their most recent devices are compared in Table I (manufacturers references). Short-term in vivo precision for total body composition has not yet been determined in these devices. However, former devices show coefficients of variation for total FM ranging from 1% to 7% depending on the devices and the population. 3,18,19 This in vivo precision is slightly higher in regional measurements. 20 Table II shows the studies that compared body composition obtained by devices of the same brand using various softwares and of different brands The differences may be due to different calibration standards 32 and different assumptions about fat distribution not revealed by the constructors. 33 Tothill et al. showed that increased adiposity led to larger differences in truncal fat, which emphasized interdevice differences between a Hologic-QDR 1000W, a Lunar DPX, and two Norland XR The same group also noticed that the proportion of bone to LBM varied in nonobese subjects: 4.8% to 7.3% for Norland, 4.2% to 5.3% for Hologic, and 4.9% to 6.2% for Lunar. 22 BODY COMPOSITION BETWEEN DXA AND REFERENCE METHODS The accuracy of the latest DXA devices for measuring total and regional body compositions has not yet been determined, neither by chemical analysis nor by comparison with in vivo reference methods. Therefore, the studies mentioned below refer mostly to former DXA devices. Multicompartment and Two-Compartment Reference Methods The DXA devices of the three manufacturers slightly overestimate FM and underestimate LBM compared with multicompartment models in healthy and ill 14,37,38 subjects. Compared with underwater weighing, Norland DXA overestimates the percentage of FM by 7.8% to 11.4%, 33,39 but Hologic and Lunar DXA give accurate measurements in healthy young subjects. 40,41 Compared with potassium counting, DXA FM is overestimated by 5.3% in postpartum women but underestimated by 0.5% in obese women. 14,42 These results demonstrate the need for more research on the accuracy of body composition methods. In clinics, assessment of changes in body composition helps evaluate the efficacy of treatments affecting nutrition status. 43 Nevertheless, the accuracy of DXA for detecting changes in FM and LBM is still under investigation. In 19 patients who lost a mean of 4 kg in 12 wk, Norland DXA accurately calculated changes in body weight compared with high-precision scales. However, the correlations between changes in body composition as assessed by DXA and a four-compartment model were worse than the intermethod correlations at baseline (r 0.33 versus 0.84 for FM and 0.45 versus 0.94 for LBM). 44 In 27 obese women undergoing a 16-wk intervention, Hologic DXA overestimated the decrease in percentage of FM, which was attributed to changes in LBM hydration. 45 Finally, Houtkooper et al. found high correlations of FM and percentage of FM changes over 1 y when comparing a Lunar DXA and a multicompartment model. 40 Overall, differences between DXA and the multicompartment model remain relatively small, ranking the DXA methodology as one of the most sensitive in detecting small changes of body composition. Magnetic Resonance Imaging (MRI) and Computed Tomography (CT) Actual reference methods for regional body composition assessment are MRI and CT. Data on comparisons with DXA are scarce. Whereas Lunar DXA and CT produce similar results for total abdominal FM, 46 Norland DXA versus MRI overestimates total abdominal FM by 50 to 125 g/cm depending on the distance from the sternal notch. 33 Several studies also tried to differentiate visceral from subcutaneous abdominal FM by combining DXA and anthropometric parameters. Jensen et al. did not confirm the accuracy of this combination and suggested instead the use of a Lunar DXA and single-slice CT. 46 However, Svendsen et al. showed that the association of total abdominal FM determined by Lunar DXA with waist-to-hip ratio and trunk skinfolds best predicted visceral abdominal FM measured by CT in 25 postmenopausal women (r 0.95). 10 Similarly, in 71 overweight subjects, the combination of abdominal transverse diameters measured by Hologic DXA with the sagittal diameter at the umbilical level and body height best correlated with visceral abdominal FM measured by CT (r 0.94 for women and 0.88 for men). 47 Although the accuracy of DXA to measure visceral abdominal FM is controversial, it may be an interesting tool to better understand the pathology and possible consequences of certain diseases and treatments. Only one study tested the accuracy of DXA for determining changes of abdominal body composition in vivo. Changes of total abdominal fat over 6 mo, measured by Norland DXA, correlated well to changes of subcutaneous fat in men and women and changes of visceral fat and total abdominal fat in women, as determined with MRI. 48 The determination of the accuracy of DXA for measuring limb muscle mass is at its very beginning. Visser et al. compared leg muscle mass derived from Hologic DXA and CT and found a good agreement between both methods for leg, thigh, and calf muscle mass, but an overestimation of DXA muscle mass in the mid-thigh region. 36 Elia et al. measured muscle mass at the mid-thigh and calf by bedside techniques and three equations derived from Hologic DXA: the first one estimated that muscle mass equaled lean soft tissue (A), the second one took into account hydration of bone (B), and the last, more complex, one was based on estimates of skin mass and on the proportion of fat in muscle and adipose tissue (C). Compared with MRI, model C showed less bias than models A and B at the mid-thigh but not at the calf. Furthermore, DXA showed better precision, a group mean closer to MRI values and smaller standard deviations of the difference compared with bedside techniques. The investigators concluded that DXA better predicted muscle mass as measured by MRI than bedside techniques. 49 REFERENCE VALUES Reference values for fat and lean body mass are essential for proper interpretation of the results. Baumgartner et al. and Rico et

49 TABLE I. CHARACTERISTICS OF THE LATEST DXA DEVICES MEASURING BODY COMPOSITION Brandnames HOLOGIC LUNAR NORLAND Types QDR-4500 Acclaim QDR-Delphi DPX-NT Prodigy XR-46 Excell Plus Material X-ray sources Alternating voltage of 2 energies Constant potential (76 kv) Constant potential (100 kv) 2-energies differentiation K-edge filter K-edge filter Effective photon energies 100 and 140 kev 38 kev and 70 kev 49 kev and 80 kev Scanning geometry Fan-beam Pencil-beam Narrow fan-beam Pencil-beam Characteristic of detector Not energy-discriminating Energy-discriminating Energy-discriminating Composition of detector Solid-state linear array Solid-state linear array Dual sodium iodide scintillation detector Table dimension (active area) 65 cm 198 cm 65 cm 198 cm 61 cm 195 cm 60 cm 198 cm 64.8 cm 193 cm NA Weight (without computer console) 365 kg 310 kg 272 kg 272 kg 204 kg 181 kg Quality assessment Soft tissue calibration Step phantom of lucite and aluminium Plastic polyoxymethylene (40% fat), water (5% fat) Step phantom of lucite and aluminium Bone calibration Internal rotating drum Internal hydroxyapatite calibration Step phantom of aluminium and acrylic Hydroxyapatite spine phantom in epoxy resin Block phantom and spine aluminium phantom Hydroxyapatite spine phantom in epoxy resin Total body precision Bone mineral density (CV) 1% 1% 1.0% 1.0% 1.0% NA Body composition NA NA %fat: SD 1 NA NA Measurement Total body effective radiation 1 mr 1.5 mr 0.02 mrem mrem 0.1 mrem NA Scan time 3 min 6.8 min 8 min 5 min 5 min NA NA: not available. 68 Genton et al. Nutrition Volume 18, Number 1, 2002

50 Nutrition Volume 18, Number 1, 2002 Dual-Energy X-ray Absorptiometry 69 TABLE II. COMPARISONS OF TOTAL BODY COMPOSITION MEASURED BY DXA DEVICES OF DIFFERENT MANUFACTURERS Authors N subjects Compared devices BMC (%)* FM (%)* FM (kg)* LBM (kg)* Pritchard et al F M QDR-1000W (5.35) vs. DPX (3.4) Tothill et al. 21 6F, 5M QDR-1000W (EWB V5.51P) vs. DPX (3.4, 3.6) NA 3.7 NA NA QDR-1000W (EWB V5.51P) vs. XR-36 (2.2.5, 2.4) NA 6.3 NA NA DPX (3.4, 3.6) vs. XR-36 (2.2.5, 2.4) NA 2.6 NA NA Abrahamsen et al F, 4M QDR-2000 pencil-beam (4.42) vs. QDR-2000 fan-beam (4.55) QDR-1000W (4.42) vs. QDR-2000 pencil-beam (4.42) Pierson et al F, 99M DPX (3.6) vs. XR-26 (2.4) 7.9 to to 6.9 NA NA Platon et al. 24 1F, 5M DPX (3.6) of two different locations Spector et al F, 11M QDR-2000 pencil-beam (EWB 5.54) vs. QDR-2000 fan-beam (EWB 5.54) QDR-2000 pencil-beam (WB v 5.54) vs. QDR-2000 pencil-beam (EWB 5.54) Van Loan et al F DPX (3.4) vs. DPX (3.6) Modlesky et al M QDR-1000W (EWB) vs. DPX-L (1.3z) Kistorp et al F QDR-2000 fan-beam (5.54 A) vs. DPX (3.6y) Oldroyd et al. 29 6F, 8M DPX (1.3y) vs. DPX-L (3.6y) Ellis et al F, 21M QDR-2000 pencil-beam (WB V5.71) vs. QDR-4500 (EWB V8.21a:3) *: result of the first device minus the second device mentioned under compared devices : these papers also studied regional body composition differences between devices F, females; M, males; WB, whole body; EWB, enhanced whole body; BMC, bone mineral content; BMD, bone mineral density; FM, fat mass; LBM, lean body mass excluding bone; NA, not available. QDR manufactured by Hologic, DPX by Lunar, and XR by Norland al. published reference values based on, respectively, 128 elderly Northern Americans using Lunar DXA 51 and 815 spanish subjects aged u15 to 83 y using Norland DXA. 51 Recently, Kyle et al. reported FM and LBM percentiles of more than 5000 healthy Swiss adults of all ages, using bioelectric impedance analysis validated, previously against DXA (Hologic QDR-4500). 52 However, bioelectric impedance analysis measurements may not correlate well with the results of all DXA devices because FM and LBM may differ between DXA devices. Furthermore, body composition is specific to a population and therefore these values cannot be used as universal reference values. Standardized calibration procedures between DXA devices may provide more detailed information on ethnic body composition differences. Finally, the health implications of body composition values need further research. CONCLUSION DXA precisely assesses total and regional body compositions, but differences between devices limit longitudinal and cross-sectional comparisons between subjects and populations. Furthermore, the accuracy of total and regional body compositions measured by DXA compared with reference methods can still be improved. From a clinical point of view, research should continue to focus on the relationship between health, disease, and body composition. REFERENCES 1. Kyle UG, Pichard C. Dynamic assessment of fat-free mass during catabolism. Curr Opin Clin Nutr Metab Care 2000;3: Heymsfield SB, Nunez C, Testolin C, Gallagher D. Anthropometry and methods of body composition measurement for research and field application in the elderly. Eur J Clin Nutr 2000;54:S26 3. Jebb SA. Measurement of soft tissue composition by dual energy X-ray absorptiometry. Br J Nutr 1997;77: Nord RH. DXA body composition properties: inherent in the physics or specific to scanner type? Appl Radiat Isot 1998;49: Kelly TL, Berger N, Richardson TL. DXA body composition: theory and practice. Appl Radiat Isot 1998;49: Pietrobelli A, Formica C, Wang Z, Heymsfield SB. Dual-energy X-ray absorptiometry body composition model: review of physical concepts. Am J Physiol 1996;271:E Blake GM. Appendix: mathematical background for body composition studies. In: Blake GM, Wahner HW, Fogelman I, eds. The evaluation of osteoporosis: dual energy X-ray absorptiometry and ultrasound in clinical practice, 2nd ed. London: Martin Dunitz,1999: Madden AM, Morgan MY. The potential use of dual-energy X-ray absorptiometry in the assessment of body composition in cirrhotic patients. Nutrition 1997;13:40 9. Laskey MA. Dual-energy X-ray absorptiometry and body composition. Nutrition 1996;12: Svendsen OL, Haarbo J, Hassager C, Christiansen C. Accuracy of measurements of body composition by dual-energy x-ray absorptiometry in vivo. Am J Clin Nutr 1993;57: Toda Y, Segal N, Toda T, Kato A, Toda F. A decline in lower extremity lean body mass per body weight is characteristic of women with early phase osteoarthritis of the knee. J Rheumatol 2000;27: Formica CA. Total body bone mineral and body composition by absorptiometry. In: Blake GM, Wahner HW, Fogelman I, eds. The evaluation of osteoporosis: dual energy X-ray absorptiometry and ultrasound in clinical practice, 2nd ed. London: Martin Dunitz,1999: Pietrobelli A, Wang Z, Formica C, Heymsfield SB. Dual-energy X-ray absorptiometry: fat estimation errors due to variation in soft tissue hydration. Am J Physiol 1998;274:E Hendel HW, Gotfredsen A, Anderson T, Hojgaard L, Hilsted J. Body composition during weight loss in obese patients estimated by dual-energy X-ray absorptiometry and by total body potassium. Int J Obes 1996;20: Tataranni PA, Ravussin E. Use of dual-energy X-ray absorptiometry in obese individuals. Am J Clin Nutr 1995;62: Hans DB, Genton L, Gonicella G, et al. Half-body dual X-rax absorptiometry predicts whole body composition (WBC): a potential method to measure obese patients. Clin Nutr 2001;20:1

51 70 Genton et al. Nutrition Volume 18, Number 1, Westmacott C, Oldroyd B, Truscott J, et al. A pilot study to measure changes in body composition in very obese women on hypocaloric diets. Appl Radiat Isot 1998;49: Slosman DO, Casez JP, Pichard C, et al. Assessment of whole-body composition using dual X-ray absorptiometry. Radiology 1992;185: Tothill P, Avenell A, Reid DM. Precision and accuracy of measurements of whole-body bone mineral: comparisons between Hologic, Lunar and Norland dual-energy X-ray absorptiometers. Br J Radiol 1994;67: Mazess RB, Barden HS, Bisek JP, Hanson J. Dual-energy X-ray absorptiometry for total-body and regional bone-mineral and soft-tissue composition. Am J Clin Nutr 1990;51: Pritchard JE, Nowson CA, Strauss BJ, et al. Evaluation of dual energy X-ray absorptiometry as a method of measurement of body fat. Eur J Clin Nutr 1993;47: Tothill P, Avenell A, Love J, Reid DM. Comparisons between Hologic, Lunar and Norland dual-energy X-ray absorptiometers and other techniques used for whole-body soft tissue measurements. Eur J Clin Nutr 1994;48: Abrahamsen B, Gram J, Hansen TB, Beck-Nielsen H. Cross calibration of QDR-2000 and QDR-1000 dual-energy X-ray densitometers for bone mineral and soft-tissue measurements. Bone 1995;16: Pierson RN Jr, Wang J, Thornton JC, et al. Bone mineral and body fat measurements by two absorptiometry systems: comparisons with neutron activation analysis. Calcif Tissue Int 1995;56: Platon NI, Macallan DC, Jebb SA, Pazianas M, Griffin GE. Dual-energy X-ray absorptiometry results differ between machines. Lancet 1995;346: Spector E, LeBland A, Shackelford L. Hologic QDR-2000 whole-body scans: a comparison of three combinations of scan modes and analysis software. Osteoporos Int 1995;5: Van Loan MD, Keim NL, Berg K, Mayclin PL. Evaluation of body composition by dual-energy X-ray absorptiometry and two different software packages. Med Sci Sports Exerc 1995;27: Modlesky CM, Lewis RD, Yetman KA, et al. Comparison body composition and bone mineral measurements from two DXA instruments in young men. Am J Clin Nutr 1996;64: Kistorp CN, Svendsen OL. Body composition analysis by dual energy X-ray absorptiometry in female diabetics differ between manufacturers. Eur J Clin Nutr 1997;51: Oldroyd B, Truscott JG, Woodrow G, et al. Comparison of in-vivo body composition using two Lunar dual-energy X-ray absorptiometers. Eur J Clin Nutr 1998;52: Ellis KJ, Shypailo RJ. Bone mineral and body composition measurements: cross-calibration of pencil-beam and fan-beam dual-energy X-ray absorptiometers. J Bone Miner Res 1998;13: Tothill P. Dual-energy X-ray absorptiometry for the measurement of bone and soft tissue composition. Clin Nutr 1995;14: Tothill P, Han TS, McNeil G, Reid DM. Comparisons between fat measurements by dual-energy X-ray absorptiometry, underwater weighing and magnetic resonance imaging in healthy women. Eur J Clin Nutr 1996;50: Goran MI, Toth MJ, Poehlman ET. Assessment of research-based body composition techniques in healthy elderly men and women using the 4-compartment model as a criterion method. Int J Obes 1998;22: Wang ZM, Deurenberg P, Guo SS, et al. Six-compartment body composition model: inter-method comparisons of total body fat measurement. Int J Obes 1998;22: Visser M, Fuerst T, Lang T, Salamone L, Harris T. Validity of fan-beam dual-enery X-ray absorptiometry for measuring fat-free mass and leg muscle mass. Health, Aging, and Body Composition Study-Dual-Energy X-ray Absorptiometry and Body Composition Working Group. J Appl Physiol 1999;87: Kemink SA, Frijns JT, Hermus AR, et al. Body composition determined by six different methods in women bilaterally adrenalectomized for treatment of Cushing s disease. J Clin Endocr Metab 1999;84: Haderslev KV, Staun M. Comparison of dual-energy X-ray absorptiometry to four other methods to determine body composition in underweight patients with chronic gastro-intestinal disease. Metabolism 2000;49: Fogelholm M, Kukkonen-Harjula TK, Sievanen HT, Oja P, Vuori IM. Body composition assessment in lean and normal-weight young women. Br J Nutr 1996;75: Houtkooper LB, Going SB, Sproul J, Blew RM, Lohman TG. Comparison of methods for assessing body-composition changes over 1 y in postmenopausal women. Am J Clin Nutr 2000;72: Snead DB, Birge SJ, Kohrt WM. Age-related differences in body composition by hydrodensitometry and dual-energy X-ray absorptiometry. J Appl Physiol 1993; 74: Butte NF, Hopkinson JM, Ellis KJ, Wong WW, Smith EO. Changes in fat-free mass and fat mass in postpartum women: a comparison of body composition models. Int J Obes 1997;21: Pichard C, Genton L, Jolliet P. Measuring body composition:a landmark of quality control for nutritional support services. Curr Opin Clin Nutr Metab Care 2000;3: Podenphant J, Gotfredsen A, Engelhart M, et al. Comparison of body composition by dual-energy X-ray absorptiometry to other estimates of body composition during weight loss in obese patients with rheumatoid arthritis. Scand J Clin Lab Invest 1996;56: Evans EM, Saunders MJ, Spano MA, et al. Body-composition changes with diet and exercise in obese women: a comparison of estimates from clinical methods and a 4-component model. Am J Clin Nutr 1999;70:5 46. Jensen MD, Kanaley JA, Reed JE, Sheedy PF. Measurement of abdominal and visceral fat with computed tomography and dual-energy X-ray absorptiometry. Am J Clin Nutr 1995;61: Bertin E, Marcus C, Ruiz JC, Eschard JP, Leutenegger M. Measurement of visceral adipose tissue by DXA combined with anthropometry in obese humans. Int J Obes Relat Metab Disord 2000;24: Kamel EG, McNeill G, Van Wijk MC. Change in intra-abdominal adipose tissue volume during weight loss in obese men and women: correlation between magnetic resonance imaging and anthropometric measurements. Int J Obes Relat Metab Disord 2000;24: Elia M, Fuller NJ, Hardingham CR, et al. Modeling leg sections by bioelectrical impedance analysis, dual-energy x-ray absorptiometry, and anthropometry: assessing segmental muscle volume using magnetic resonance imaging as a reference. Ann NY Acad Sci 2000;904: Baumgartner RN, Stauber PM, McHugh D, Koehler KM, Garry PJ. Crosssectional age differences in body composition in persons 60 years of age. J Gerontol 1995;50A:M Rico H, Revilla M, Villa LF, et al. The four-compartment models in body composition: data from a study with dual-energy X-ray absorptiometry and near-infrared interactance on 815 normal subjects. Metabolism 1994;43: Kyle UG, Genton LC, Slosman DO, Pichard C. Fat-free and fat mass percentiles in 5225 healthy subjects aged 15 to 98 years. Nutrition 2001;17:534

52 APPENDIX 2

53 ARTICLE IN PRESS Clinical Nutrition (2006) 25, ORIGINAL ARTICLE Comparison of body weight and composition measured by two different dual energy X-ray absorptiometry devices and three acquisition modes in obese women Laurence Genton a,b,,véronique L. Karsegard b, Sophie Zawadynski a, Ursula G. Kyle b, Claude Pichard b, Alain Golay c, Didier B. Hans a a Nuclear Medicine, Geneva University Hospital, 1211 Geneva 14, Switzerland b Clinical Nutrition, Geneva University Hospital, 1211 Geneva 14, Switzerland c Therapeutic Education for Chronic Diseases, Geneva University Hospital, 1211 Geneva 14, Switzerland Received 19 May 2005; accepted 4 November 2005 KEYWORDS Body composition; DXA; Inter-device differences; Reproducibility; Obesity Summary Background & aims: Weight measured by dual-energy X-ray (DXA) was shown to be increasingly underestimated in subjects over 75 kg compared to an electronic scale. This study compares body weight and composition measured by balance beam scale and three DXA acquisition modes in obese subjects. Methods: In 39 obese, body weight was measured by balance beam scale, and body weight and composition by DXA Hologic QDR4500A s in normal (NPM) and high power mode (HPM) (Enhanced v8.26 and v8.26* softwares) and DXA GE-Lunar Prodigy s (v6.5 software). To ensure linearity of body weight and composition measured by the different DXA acquisitions, we also measured 13 women with a body mass index (BMI) of kg/m 2. Results: While QDR4500A HPM overestimates scale weight by about 2 kg over the whole BMI spectrum, QDR4500A NPM underestimates scale weight as a weightdependent response ( kg overall, kg in morbidly obese women). These results suggest switching from one mode to the other at a specific threshold, i.e. in our study a weight of 90 kg or a BMI of 34 kg/m 2. Prodigy gives weight about similar to scale ( kg). Both Hologic acquisition modes underestimate fat mass but overestimate lean body mass compared to Prodigy. Conclusions: The QDR4500A NPM is inappropriate in women over 90 kg. Unfortunately, the QDR4500A HPM overestimates body weight in the range of kg. The Corresponding author. Clinical Nutrition, Geneva University Hospital, 1211 Geneva 14, Switzerland. Tel.: ; fax: address: laurence.genton@hcuge.ch (L. Genton) /$ - see front matter & 2005 Elsevier Ltd and European Society for Clinical Nutrition and Metabolism. All rights reserved. doi: /j.clnu

54 ARTICLE IN PRESS Body composition with two DXA devices obese women 429 difference between scale and Prodigy weight remains stable throughout weight ranges. To better assess their accuracies in terms of body composition, QDR4500A NPM, HPM and Prodigy should be tested against phantoms or in vivo multicompartment models. & 2005 Elsevier Ltd and European Society for Clinical Nutrition and Metabolism. All rights reserved. Introduction In clinical practice, practitioners use body mass index (BMI) as an estimation of fat mass (FM) or determine body composition by bedside techniques such as bio-impedance analysis or skinfold measurement. The formulae used to derive fat and lean body (LBM) masses from these techniques are validated either against multi-compartment models or against a single method such as dual-energy X-ray absorptiometry (DXA), hydrostatic weighing, total body potassium or in vivo neutron activation analysis. 4 DXA is the preferred technique to measure whole body bone mineral content (BMC), FM and LBM. It is however not yet considered a gold standard for determining body composition. In some studies, it overestimates FM and underestimates LBM in healthy and ill subjects compared to multi-compartment models. 1,2 Compared to underwater weighing, Hologic and Lunar DXA give accurate measurements in healthy young subjects. Compared to total body potassium counting, DXA FM is overestimated by 5.3% in post-partum women, but underestimated by 0.5% in obese women. These results demonstrate the need for further research on the accuracy of body composition methods. 4 Furthermore, the results differ according to the devices used. Presently, the three most common DXA manufacturers are Hologic Inc., GE-Lunar Inc. and Norland Medical Systems. While their most recent devices have not been compared, former Hologic devices measured lower % FM than the Lunar ( 3.7%) and Norland devices ( 6.3%). 3 Body composition measurements also slightly vary between older and newer devices within the same company because of changes in the underlying technology and software. 4,5 For instance, the QDR 4500 A s by Hologic underestimates %FM compared to the older QDR 2000 s, 6 as does the DPX-L s of Lunar relative to the older DPX s. 7 These inter- and intra-manufacturer differences demonstrate the need of constant validation of new DXA devices in vitro and/or in vivo against multi-compartment models. However, our preliminary work has shown that an in vivo validation is difficult to perform in obese people because they often do not fit in the scanning area. 8 Besides technology, the thickness of the measured subjects also influences DXA results. In vitro, DXA underestimates fat tissue compared to direct chemical analysis at a meat thickness up to 26 cm. At a thickness over 26 cm, fat measured by Hologic QDR 1000 s increased markedly. 9 Economos et al. 10 used phantoms combined with various thicknesses of soft-tissue overlays and noted that bone mineral content measured by Lunar DPX s, Norland XR s and Hologic QDR 2000 s decreased as thickness increases from 20.5 to 26.0 cm. While the in vivo effect of abdominal thickness on body composition measured by recent DXA devices is yet unknown, a former abstract showed that weight, measured by Hologic QDR 4500 A s in normal mode, was increasingly underestimated in subjects over 75 kg compared to an electronic scale. 11 The high power mode software has been designed especially for obese subjects, in order to correct this underestimation. We hypothesized that the measurements of body weight and composition differ between the DXA devices and acquisitions, and that these differences appear predominantly in obese subjects. The study aim was to compare the body weight and composition measured by three different DXA acquisition modes, the Hologic QDR 4500 A s in normal and high power mode and the Lunar Prodigy s, in obese women (body mass index (BMI) X30 kg/m 2 ). To ensure that the mode of acquisitions has little influence on body composition of non-obese subjects, and to ensure the linearity of our results, these measurements were also performed in subjects with a BMI between 25.0 and 29.9 kg/m 2. To our knowledge, this is the first time that body composition and weight between these DXA acquisitions have been compared in obese subjects. Subjects and methods This study included 39 Caucasian women with a BMI X30 kg/m 2 (BMI ¼ weight (kg)/height 2 (m)) who were hospitalized at the Geneva University Hospital for a future gastroplasty or recruited at the outpatient obesity clinics. To ensure the linearity of body weight and composition measured by

55 ARTICLE IN PRESS 430 different DXA acquisitions in lower BMI ranges, we included additional 13 healthy Caucasian women with a BMI between 25 and 29.9 kg/m 2. All subjects signed a written informed consent. This protocol was accepted by the Ethical Committee of our institution. Just before performing the DXA measurements and while the subjects were in underwear and without shoes, the subjects body height was measured to the nearest 0.5 cm with a height gauge and their body weight to the nearest 0.1 kg on a balance beam scale (Seca, Germany) crosscalibrated using known calibration weights. BMC, FM and soft-tissue LBM were determined by DXA QDR 4500-A s (Hologic Inc., Waltham, MA, USA), and DXA Prodigy s (GE-Lunar, Madison, WI, USA). The QDR 4500-A scans were performed in normal (NPM) and high power mode (HPM) with the Enhanced 8.26 and 8.26* Whole-body software versions, respectively. The Prodigy acquisition was performed systematically in thick mode (v6.5 whole body software). To facilitate the reading, the DXA acquisitions will be termed QDR 4500-A NPM, QDR 4500-A HPM and Prodigy in this paper. To calculate short-term precision of body weight and composition using the QDR 4500-A HPM and Prodigy, a subgroup of subjects underwent the same DXA acquisition twice (n ¼ 10 for QDR 4500-A NPM and n ¼ 14 for QDR 4500-A HPM and Prodigy) within 1 h and with repositioning. All measurements were performed according to the manufacturers protocols. In addition, long-term stability of both devices was checked and assured by daily quality control using the phantoms provided by the each respective manufacturer. These DXA devices are based on the same physical principle but differ in their conception. 4 Both use an X-ray tube to direct two energy levels of photons through the body. In the QDR 4500-A NPM and HPM, the X-ray generator emits switched pulsed radiation of 100 and 140 kv. The difference between the two modes is only related to the intensity of the emitted flux, which is 5 ma for the QDR 4500-A NPM and 10 ma for the HPM. The Prodigy uses a constant potential of 76 kv (0.150 ma) and a K-edge filter to discriminate photon energies of 38 and 70 kev. The acquisition time also differs between the two Hologic modes and the Prodigy. 4 Scanning is performed in fanbeam-mode for the QDR 4500-A NPM and HPM and narrow fan-beam mode for the Prodigy. As they pass through the body, these two X-ray beams are attenuated due to the absorption and scattering of the photons. The attenuation is measured for every pixel of the body surface by a detector, whose composition is similar between both devices. Complex calculations allow determination of bone mineral and soft tissue densities. Soft tissue can further be separated into FM and LBM as they have different attenuation characteristics. 4 Statistical analysis Patients were analyzed altogether and separately according to BMI categories (25 30, 30 35, and X40 kg/m 2 ). Results are expressed as mean7standard deviation. Continuous data were compared with unpaired t-tests or with analysis of variance, followed by Fisher s post hoc test in case of significance. Simple regressions tested the relationship between body weight measured by balance beam scale and DXA. Body composition between DXA acquisitions was compared by the analysis of Bland and Altman. 12 Short-term in vivo precision was calculated from duplicate measurements with repositioning as follows 13 : vffiffiffiffiffiffiffiffiffiffiffiffiffiffi up m RMS CV ¼ RMS SD 100% with RMS SD ¼ x u t SD 2 j j¼1 where RMS CV corresponds to the root mean square of the coefficient of variation, x to the mean of weight, FM, LBM or BMC in the repeated measurements and SD to the standard deviation of the mean. The error of measurement with a 95% confidence interval (95% CI) was calculated as 2 O2 RMS CV. Statistical analysis was performed with Statview 5.0 (Abacus Concept Inc., Berkeley, CA). Significance was set at Po0:05. Results L. Genton et al. Fifty-two women were included in the study. Age was similar between BMI categories ( , , and yr for the categories 25 30, 30 35, and X40 kg/m 2, respectively). Women of the BMI category kg/m 2 were significantly taller ( cm) than women of the group kg/m 2 ( cm, P ¼ 0:026), kg/m 2 ( cm, P ¼ 0:022) and X40 kg/m 2 ( cm, P ¼ 0:007). Height between the other BMI categories was not significantly different. All included women underwent body composition measurements with the QDR 4500-A device, but 16 women could not be measured with the Prodigy device because they did not fit in the scanning area. Thus, we performed one analysis comparing the two QDR 4500-A acquisitions in all subjects and m,

56 ARTICLE IN PRESS Body composition with two DXA devices obese women 431 another analysis comparing the two QDR 4500-A and the Prodigy acquisition in 36 subjects. In our study, QDR 4500-A NPM underestimated and QDR 4500-A HPM overestimated weight compared to our balance beam scale, although not significantly when considering all subjects together (Table 1). In women with a BMI X40 kg/m 2, QDR 4500-A NPM measured significantly lower weight than QDR 4500-A HPM ( kg, P ¼ 0:017, or %). Regressions of scale weight against DXA weight showed that the slope improved when using the QDR 4500-A HPM instead of the QDR A NPM in all BMI categories (r ¼ 0:9821:04 vs ). In women with a BMI X40 kg/m 2 who underwent the three whole-body acquisitions, weight measured by QDR 4500-A NPM was significantly lower than weight measured by QDR 4500-A HPM ( kg, P ¼ 0:008, or %) and Prodigy ( kg, P ¼ 0:037, or %). In the same BMI category, mean weight measured by Prodigy was the closest to scale weight but the slope was better with QDR 4500-A HPM than Prodigy (r ¼ 0:91 versus 0.84). The regressions of DXA against scale weight showed an increasing underestimation of weight by QDR 4500-A NPM with increasing scale weight (Fig. 1). Regarding body composition (Table 2), QDR 4500-A HPM measured significantly higher LBM than QDR 4500-A NPM in subjects with a BMI X40 kg/m 2 ( kg, Po0:001, or %). FM and BMC were similar between acquisitions independently of BMI category. In the subjects who underwent the three whole-body acquisitions (Table 3), QDR 4500-A HPM measured higher LBM than Prodigy in the group kg/m 2 ( kg, P ¼ 0:024, or %) and than QDR 4500-A NPM ( kg, P ¼ 0:042, or %) and Prodigy ( kg, P ¼ 0:002, or %) in the group X40 kg/m 2. FM was significantly higher with Prodigy than QDR 4500-A NPM ( kg, P ¼ 0:002, or %) and HPM in women X40 kg/m 2 ( kg, P ¼ 0:014, or %). BMC was higher with Prodigy than QDR A NPM and HPM in the group kg/m 2 ( , P ¼ 0:003 and , P ¼ 0.002, respectively) and in the group X 40 kg/m 2 ( , P ¼ and , P ¼ 0:003, respectively). Bland Altmann analysis (Fig. 2) showed that, when comparing body composition of QDR 4500-A HPM and Prodigy, FM and LBM differences increased with higher weights. Short-term precision was measured in 10 subjects for the QDR 4500-A NPM (BMI: kg/m 2 ) and was 0.63% for weight (95% CI71.77%), 0.94% for LBM (95% CI72.64%), 1.04% for FM (95% CI72.92%) and 1.01% for BMC (95% CI72.84%). For the QDR 4500-A HPM and Prodigy, short-term precision was measured in 14 women (BMI: kg/m 2 ). Coefficients of variation of weight, LBM, FM and BMC were 0.25% (95% CI %), 0.63% (95% CI71.77%), 1.13% (95% CI73.18%) and 1.06% (95% CI72.99%) for QDR 4500-A HPM and 0.16% (95% CI70.45%), 1.00% (95% CI72.81%), 1.18% (95% CI73.32%) and 1.65% (95% CI74.64%) for Prodigy, respectively. Since weight measured by the Hologic acquisitions differed from scale weight, we estimated the threshold of scale weight above that we should switch from NPM to HPM. Therefore, we plotted the scale weight against the difference between scale and DXA weight, and added the 95% CI calculated previously (Fig. 3). The results showed an underestimation of weight measured by QDR 4500-A NPM when scale weight exceeded approximately 90 kg (BMI434 kg/m 2 ) and suggest that we should switch to the HPM at this point. However, the QDR 4500-A HPM overestimated weight at about 90 kg. This overestimation decreased with higher scale weight. In contrast, the weight measured by Prodigy and scale remained similar over the weight range. Discussion Body composition is an important parameter for the follow-up of obesity treatment but 40% of our subjects with a BMI over 30 kg/m 2 did not fit within the scanning area of the Prodigy table. Our preliminary data suggested that half-body scans can overcome this limitation. 8 This study highlights the difficulty of accurately measuring body composition in obese women. Comparisons of scale weight with weight measured by DXA devices In women with a BMI below 30 kg/m 2, QDR 4500-A NPM, HPM and Prodigy measure similar body weight to traditional scale. In obese women, QDR 4500-A NPM underestimates weight relative to scale, while QDR 4500-A HPM overestimates it. Only Prodigy measures weight similar to scale in all BMI categories. To our knowledge, no study has yet compared scale weight and weight measured by Hologic QDR 4500-A and Lunar Prodigy in obese subjects. Our study shows that weight measured by QDR 4500-A NPM is increasingly underestimated with higher scale weights, which confirms the results of a former abstract. 11 However, no increasing underestimation occurred with QDR 4500-A HPM and Lunar Prodigy.

57 ARTICLE IN PRESS 432 L. Genton et al. Table 1 Weight assessed by scale and both DXA devices. n Mean7SD D7SD Slope7SE r n Mean7SD D7SD Slope7SE r All Scale QDR 4500-A NPM QDR 4500-A HPM Prodigy pBMIo30 kg/m Scale QDR 4500-A NPM QDR 4500-A HPM Prodigy pBMIo35 kg/m Scale QDR 4500-A NPM QDR 4500-A HPM Prodigy pBMIo40 kg/m Scale QDR 4500-A NPM QDR 4500-A HPM Prodigy BMIX40 kg/m Scale QDR 4500-A NPM QDR 4500-A HPM y y Prodigy y D: weight of scale-weight of DXA, r: coefficient of correlation, y Po0:05 vs. QDR 4500-A NPM. Measurements performed by Lunar Prodigy and Hologic QDR 4500-A in normal (NPM) and high power (HPM) mode.

58 ARTICLE IN PRESS Body composition with two DXA devices obese women 433 Comparisons of body composition between Hologic and Lunar DXA devices Body composition (soft tissue LBM, BMC and FM) markedly differed between the QDR 4500-A (both HPM and NPM) and Prodigy. BMC was higher with Lunar Prodigy than Hologic QDR 4500 A s. These results confirm former studies, which show a systematic overestimation of BMC with Lunar devices compared to Hologic devices. 14 QDR A HPM and NPM measure lower FM and BMC and higher LBM than Prodigy at high BMIs. Interestingly, the FM measured by QDR 4500-A HPM is closer to DXA weight (kg) Scale weight (kg) Hologic NPM Hologic HPM Lunar Hologic NPM = *scale weight R 2 = Hologic HPM = *scale weight R 2 = Lunar = *scale weight R 2 = Figure 1 Scale weight plotted against DXA weight measured by the QDR 4500-A NPM, HPM and the Prodigy, and the corresponding regression equations. Weight measured by QDR 4500-A NPM clearly diverts from the line of equality. the one measured by Prodigy while the LBM measured by QDR 4500-A NPM is closer to the one measured by Prodigy. The variations in body composition may be due to differences in the manufacturer s algorithm to separate bone and soft tissue as well as slightly different levels of energy beams between QDR 4500-A HPM, NPM and Prodigy. Consequently, validation against phantoms or in vivo multi-compartment models are required in the future in order to determine which device and acquisition mode are more accurate for body composition measurements in obese women. These results may be specific to our devices. However, the tendencies should remain similar when using other devices of the same model since the DXA devices were daily calibrated with the respective phantoms provided by the manufacturers, as part of our standard operation procedures for quality control, and checked for stability over time on a day-to-day basis using the QC plot available in the software. So far, no study has compared the Hologic QDR 4500 s with the Lunar Prodigy s with regard to body composition. Three studies compared the Hologic QDR 1000W s with either the Lunar DPX s or DPX-L s in subjects of normal BMI. 3,15,16 The Hologic device underestimated FM by 1.2 to 1.4 kg and overestimated LBM by 1.4 to 2.4 kg relative to the Lunar device, while the results for BMC were inconsistent. One study compared the more recent Hologic QDR 2000 s fan-beam with the DPX s in 85 diabetic women with a BMI ranging from 18 to 43 kg/m Table 2 Body composition assessed by one DXA device with two different DXA softwares (mean7sd). n Lean body mass (kg) Bone mineral content (kg) Fat mass (kg) Fat mass (%) All 52 QDR 4500-ANPM QDR 4500-A HPM y oBMIo30 kg/m 2 13 QDR 4500-A NPM QDR 4500-A HPM oBMIo35 kg/m 2 9 QDR 4500-A NPM QDR 4500-A HPM oBMIo40 kg/m 2 9 QDR 4500-A NPM QDR 4500-A HPM BMI440 kg/m 2 21 QDR 4500-A NPM QDR 4500-A HPM y y Po0:05 vs. QDR 4500-A NPM (unpaired t-test). Measurements performed by Hologic QDR 4500 A in normal (NPM) and high power (HPM) mode.

59 ARTICLE IN PRESS 434 Table 3 Body composition assessed by both DXA devices in a subgroup of patients (mean7sd). L. Genton et al. n Lean body mass (kg) Bone mineral content (kg) Fat mass (kg) Fat mass (%) All 36 QDR4500-ANPM QDR 4500-A HPM y Prodigy z y;z y;z 25oBMIo30 kg/m 2 13 QDR 4500-A NPM QDR 4500-A HPM Prodigy z y;z z 30oBMIo35 kg/m 2 8 QDR 4500-A NPM QDR 4500-A HPM Prodigy z 35oBMIo40 kg/m 2 8 QDR 4500-A NPM QDR 4500-A HPM Prodigy z BMI440 kg/m 2 7 QDR 4500-A NPM QDR 4500-A HPM y Prodigy z y;z y;z y;z y Po0:05 vs. QDR 4500-A NPM, z vs. QDR 4500-A HPM (ANOVA). Measurements performed by Lunar Prodigy and Hologic QDR 4500 A s in normal (NPM) and high power (HPM) mode. The Hologic QDR 2000 s underestimated BMC ( 6.0%) and LBM ( 2.1 kg) and overestimated FM (+1.4 kg) relative to the DPX. In our study, we found the opposite tendency, i.e. the Hologic QDR 4500 A s in NPM and HPM measured lower FM and higher LBM than the Prodigy. Thus, body composition measurements are either different between the Hologic QDR 2000 s and QDR 4500 A s or between the Lunar DPX and Prodigy. The former hypothesis seems more plausible. Tylavski et al. 18 showed that the Hologic QDR 4500 s (software 8.21) overestimated LBM by 5.5 kg compared to the Hologic QDR 2000 s in elderly subjects with a wide range of BMI. In contrast, Mazess and Barden 19 found no significant difference of soft tissue composition between the Lunar DPX and Prodigy, in 49 adults ranging in weight from 50 to 120 kg. Comparison of Hologic and Lunar DXA devices with multicompartment models In elderly subjects with a wide range of BMI, the Hologic QDR 4500 s (software v8.21, mode not reported) overestimated body weight by 2 kg and LBM including BMC by 2.6 kg compared to a fourcompartment model. Tylavsky et al. 18 suggested applying a correction factor to the sum of LBM and BMC measured by Hologic QDR 4500 s to obtain equivalent results to the four-compartment model (LBM+BMC 4C ¼ 0.96* (LBM+BMC QDR4500 )). However, this equation did not correct for body weight, and a higher FM compensated for the lower LBM and BMC. If we assume that this correction factor can also be applied to obese women (BMI X30 kg/m 2 ), only corrected results (LBM+BMC) for QDR 4500-A NPM are similar to LBM+BMC measured by Prodigy ( vs kg) while the corrected value for the QDR 4500-A HPM still remains higher ( kg). Since Tylavsky et al. supposedly used the Hologic NPM, this extrapolation suggests that the Prodigy measures similar body composition to their multi-compartment model. Former models, such as the Lunar DPX s and DPX-L s, measured a similar LBM and FM to multi-compartment models in healthy and ill subjects with various BMIs. 1,20,21 Short-term precision In our study, precision of body composition scan was similar for both devices. While the precision of the Prodigy has not yet been assessed in other studies, the precision of the QDR 4500 s (software v8.21),

60 ARTICLE IN PRESS Body composition with two DXA devices obese women 435 FM (Prodigy QDR 4500-A NPM) (A) FM (Prodigy + QDR 4500-A NPM)/2 FM (Prodigy + QDR 4500-A HPM)/2 FM (QDR 4500-A NPM + HPM)/2 FM (Prodigy QDR 4500-A HPM) FM (QDR 4500-A NPM HPM) (B) LBM (Prodigy + QDR 4500-A NPM)/2 LBM (Prodigy + QDR 4500-A HPM)/2 LBM (QDR 4500-A NPM + HPM)/2 LBM (Prodigy QDR 4500-A NPM) LBM (Prodigy QDR 4500-A HPM) LBM (QDR 4500-A NPM HPM) (C) BMC (Prodigy + QDR 4500-A NPM)/2 BMC (Prodigy + QDR 4500-A HPM)/2 BMC (QDR 4500-A NPM + HPM)/2 BMC (Prodigy QDR 4500-A NPM) BMC (Prodigy QDR 4500-A HPM) BMC (QDR 4500-A NPM HPM) Figure 2 Bland Altman analysis for fat mass (FM) (A), lean body mass (LBM) (B) and bone mineral content (BMC) (C), between the QDR 4500-A NPM, HPM and the Prodigy. The figure shows the limits of agreement calculated as mean difference72sd. calculated by Cordero-MacIntyre et al. 22 via duplicate scans of 9 obese women, was 1.2% for FM, 1.3% for LBM and 2.1% for BMC for software V8.21a, thus slightly worse than our results. Our results suggest that in women above 90 kg (BMI434 kg/m 2 ), the use of the QDR 4500-A NPM is no longer appropriate. Unfortunately, the proposed alternative (QDR 4500-A HPM) seems to overestimate the body weight in the range of kg. Since the intensity of the flux doubles between the NPM and HPM (5 ma versus 10 ma), the use of a HPM at an intensity of 7.5 ma could be more accurate for obese women. The current thick mode of Prodigy overcomes the problem encountered in obese women as the difference between scale and DXA weights remains stable throughout the weight ranges. Our results also raise the question of whether the weight or the BMI should be taken as threshold for changing from NPM to HPM. Indeed, a few patients with a BMI lower than 34 kg/m 2 had a weight above 90 kg. To better compare and assess their respective accuracies in terms of body composition (FM, LBM and BMC), both QDR 4500-A and Prodigy should be

61 ARTICLE IN PRESS 436 L. Genton et al (A) Scale weight (kg) Weight (QDR 4500-A NPM - scale ) in % Weight (QDR 4500-A HPM - scale) in % Scale weight (kg) Weight (Prodigy - scale) in % Scale weight (kg) (B) Body mass index (kg/m 2 ) Body mass index (kg/m 2 ) Body mass index (kg/m 2 ) Weight (QDR 4500-A NPM - scale) in % Weight (QDR 4500-A HPM - scale) in % Weight (Prodigy - scale) in % Figure 3 Scale weight (A) and body mass index (B) plotted against the DXA weight expressed as % of scale weight. The technical error at 95% confidence intervals, calculated with our data on reproducibility of DXA weights, are drawn on each side of the 0 line. Arrows show the weight and body mass index at which point the value measured by QDR 4500-A NPM exceeds the technical error related to the device. tested against phantoms or in vivo multi-compartment models. Acknowledgements The authors thank the Foundation Nutrition 2000Plus for financial support as well as Giulio Conicella and Nadine Maisonneuve who helped for the measurements of body composition and Vincent Wazner who helped for scientific input. References 1. Wang ZM, Deurenberg P, Guo SS, et al. Six-compartment body composition model: inter-method comparisons of total body fat measurement. Int J Obes Relat Metab Disord 1998;22: Haderslev KV, Staun M. Comparison of dual-energy X-ray absorptiometry to four other methods to determine body composition in underweight patients with chronic gastrointestinal disease. Metabolism 2000;49: Tothill P, Avenell A, Love J, Reid DM. Comparisons between Hologic, Lunar and Norland dual-energy X-ray absorptiometers and other techniques used for whole-body soft tissue measurements. Eur J Clin Nutr 1994;48: Genton L, Hans D, Kyle UG, Pichard C. Dual-energy X-ray absorptiometry and body composition: differences between devices and comparison with reference methods. Nutrition 2002;18: Nakata Y, Tanaka K, Mizuki T, Yoshida T. Body composition measurements by dual-energy X-ray absorptiometry differ between two analysis modes. J Clin Densitom 2004;7: Ellis KJ, Shypailo RJ. Bone mineral and body composition measurements: cross-calibration of pencil-beam and fanbeam dual-energy X-ray absorptiometers. J Bone Miner Res 1998;13: Oldroyd B, Truscott JG, Woodrow G, et al. Comparison of invivo body composition using two Lunar dual-energy X-ray absorptiometers. Eur J Clin Nutr 1998;52: Genton L, Hans D, Kyle UG, Slosman DO, Pichard C. Halfbody dual-energy X-ray absorptiometry (DXA) predicts whole body composition (WBC) in obese patients. Clin Nutr 2003; 21: Jebb SA, Goldberg GR, Jennings G, Elia M. Dual-energy X-ray absorptiometry measurements of body composition: effects of depth and tissue thickness, including comparisons with direct analysis. Clin Sci (Lond) 1995;88:

62 ARTICLE IN PRESS Body composition with two DXA devices obese women Economos CD, Nelson ME, Fiatarone Singh MA, et al. Bone mineral measurements: a comparison of delayed gamma neutron activation, dual-energy X-ray absorptiometry and direct chemical analysis. Osteoporos Int 1999;10: Sutter B, Legrand O, Meys E, Bougon F, Hardouin P. Accuracy in weight mesurement with Hologic QDR 4500 Acclaim. Bone 2001;28:S Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1: Njeh CF, Hans D. Instrument evaluation. In: Njeh CF, Langton CM, editors. The physical measurement of bone. London: Institute of Physics Publishing; p Laskey MA, Flaxman ME, Barber RW, et al. Comparative performance in vitro and in vivo of Lunar DPX and Hologic QDR-1000 dual energy X-ray absorptiometers. Br J Radiol 1991;64: Pritchard JE, Nowson CA, Strauss BJ. Evaluation of dual energy X-ray absorptiometry as a method of measurement of body fat. Eur J Clin Nutr 1993;47: Modlesky CM, Lewis RD, Yetman KA, et al. Comparison of body composition and bone mineral measurements from two DXA instruments in young men. Am J Clin Nutr 1996;64: Kistorp CN, Svendsen OL. Body composition analysis by dual energy X-ray absorptiometry in female diabetics differ between manufacturers. Eur J Clin Nutr 1997;51: Tylavsky F, Lohman T, Blunt BA, et al. QDR 4500A DXA overestimates fat-free mass compared with criterion methods. J Appl Physiol 2003;94: Mazess RB, Barden HS. Evaluation of differences between fan-beam and pencil-beam densitometers. Calcif Tissue Int 2000;67: Goran MI, Toth MJ, Poehlman ET. Assessment of researchbased body composition techniques in healthy elderly men and women using the 4-compartment model as a criterion method. Int J Obes Relat Metab Disord 1998;22: Kemink SA, Frijns JT, Hermus AR, Pieters GF, Smals AG, van Marken Lichtenbelt WD. Body composition determined by six different methods in women bilaterally adrenalectomized for treatment of Cushing s disease. J Clin Endocrinol Metab 1999;84: Cordero-MacIntyre ZR, Peters W, Libanati CR, et al. Reproducibility of DXA in obese women. J Clin Densitom 2002;5:35 44.

63 APPENDIX 3

64 APPLIED NUTRITIONAL INVESTIGATION Single Prediction Equation for Bioelectrical Impedance Analysis in Adults Aged Years Ursula G. Kyle, MS, RD, Laurence Genton, MD, Laurie Karsegard, MS, Daniel O. Slosman, MD, and Claude Pichard, MD, PhD From the Divisions of Clinical Nutrition and Dietetics and Nuclear Medicine, Geneva University Hospital, Geneva, Switzerland Existing equations for bioelectrical impedance analysis (BIA) are of limited use when subjects age or become overweight because these equations were developed in young, normal-weight subjects and are not valid in elderly or overweight people. The purpose of this study was to validate a single BIA equation in healthy white subjects aged y with a body mass index between 17.0 and 33.8 kg/m 2. Healthy subjects (202 men and 141 women) aged y were measured by two methods: fat-free mass (FFM) was measured by dual-energy x-ray absorptiometry (Hologic QDR-4500) and by a bioelectrical impedance analyzer (Xitron 4000B). Validity of BIA was assessed by double cross validation. Because correlations were high (r ) and prediction errors low, a single equation was developed using all subjects, as follows: FFM (0.518 height 2 /resistance) (0.231 weight) (0.130 reactance) (4.229 sex: men 1, women 0). FFM predicted with dual-energy x-ray absorptiometry was kg. BIA-predicted FFM was kg (r 0.986, standard error of the estimate 1.72 kg, technical error 1.74 kg). In conclusion, the new Geneva BIA equation was valid for prediction of FFM in healthy adults aged y with body mass indexes between 17.0 and 33.8 kg/m 2. Inclusion of reactance in the single prediction equation appeared to be essential for use of BIA equations in populations with large variations in age or body mass. Nutrition 2001;17: Elsevier Science Inc Key words: dual-energy x-ray absorptiometry, fat-free mass, bioelectrical impedance analysis, body composition INTRODUCTION Assessment of fat-free mass (FFM) and fat mass (FM) in patients optimizes nutrition support to avoid or minimize muscle wasting or obesity. Nutrition assessment should therefore include objective body-composition measurements. Bioelectrical impedance analysis (BIA) has been developed for field use and has shown great potential for use in estimating body composition; it is also easy to use, non-invasive, and inexpensive. 1,2 Many investigators have developed empiric BIA equations for prediction of FFM, total body weight, and body fat. 1,3 13 Most of these equations have been validated in relatively young, healthy adults against several body-composition techniques. 9 Studies have shown that BIA formulas developed for healthy, normal-weight subjects are not suitable for obese subjects 14,15 and are not valid in elderly subjects. 13 In longitudinal studies, the use of different BIA formulas in the same subject who becomes overweight or ages introduces a bias into body-composition studies and makes one question whether the differences in body composition are due to changes in the BIA formula or to changes in body composition. Thus, it would be advantageous to use a single formula that is applicable in young and elderly subjects and permits estimation of FFM and FM in overweight subjects and subjects with grade 1 obesity. This study was supported by Foundation Nutrition 2000Plus. Correspondence to: Claude Pichard, MD, PhD, Head, Clinical Nutrition and Diettherapy, Geneva University Hospital, 1211 Geneva, Switzerland. pichard@cmu.unige.ch Date accepted: September 26, Roubenoff et al. 13 and others concluded that BIA equations are subject to errors that cannot be determined a priori unless they are validated in the specific population in which they are to be applied Thus, BIA equations must be validated in a representative population sample against a reference method before it can be accepted as accurate. BIA can be validated against dualenergy x-ray absorptiometry (DXA), hydrodensitometry, and total body potassium. DXA is one reference method 3 that has been validated against independent methods such as in vivo neutron activation, 5,6 total body potassium, and hydrodensitometry. 7 Recent studies have shown that DXA and hydrodensitometry agree well at high, moderate, and low levels of body fat. 19 The purpose of this study was to validate, against DXA, a single BIA equation in a healthy, white European group of subjects aged y and with body mass indexes (BMIs) of kg/m 2. A single BIA equation that is valid in subjects with different ages and BMIs for use in longitudinal studies would be a significant advantage for nutrition assessment. SUBJECTS AND METHODS Subjects Three hundred forty-three healthy ambulatory white subjects (202 men and 141 women) aged y (Table I) were included in this study. Subjects were non-randomly recruited through advertisements in local newspapers, and invitations to participate in the study were sent to elderly members of leisure clubs. Although subjects were non-randomly selected, statistical analysis showed no difference in height, weight, and BMI between subjects and a control group of healthy age-matched 1411 men and 1759 women. Nutrition 17: , /01/$20.00 Elsevier Science Inc., Printed in the United States. All rights reserved. PII S (00)

65 Nutrition Volume 17, Number 3, 2001 Fat-Free Mass and Bioelectrical Impedance 249 TABLE I. ANTHROPOMETRIC AND BIA CHARACTERISTICS OF HEALTHY SUBJECTS Age (y) Men n Height (cm) Weight (kg) BMI (kg/m 2 ) IBW (%) Resistance ( ) Reactance ( ) H 2 /R Women n Height (cm) Weight (kg) IBW (%) BMI (kg/m 2 ) Resistance ( ) Reactance ( ) H 2 /R BIA, bioelectrical impedance analysis; BMI, body mass index; H 2 /R, height squared/resistance; IBW, ideal body weight. Exclusion criteria were active medical treatment, hospitalization within 3 mo of measurement, or a physical handicap that might interfere with body-composition measurement (e.g., amputation or paralysis). Body height was measured to the nearest 0.5 cm and body weight was measured to the nearest 0.1 kg on a balance-beam scale. Heights and weights of both groups were normally distributed. Percentage of ideal body weight was derived from 1983 Metropolitan Life Insurance Tables. Values used were the midpoints for medium frame for each height, recalculated in centimeters and kilograms. Three women older than 65 y had BMIs below 18.0 kg/m 2, but their medical histories showed no recent weight loss or illness, so they were considered healthy. Each subject was measured by BIA and then by DXA. All subjects signed informedconsent statements and the study protocol was approved by the Geneva University Hospital Ethics Committee. Bioelectrical Impedance Analysis Briefly, a 5-kHz, 0.8-mA electric current was produced by a Xitron 4000B generator (Xitron Technologies, San Diego, CA, USA) and applied to the skin using adhesive electrodes (3 M Red Dot, 3 M Health Care, Borken, Germany) with subjects in the supine position. Standard procedures were used. 5 Resistance and reactance were measured by the BIA generator and used to mathematically derive FFM, as previously described, 3,5 using the formula V H 2 /R, where conductive volume (V) represents FFM, is the specific resistivity of the conductor, height (H) is the length of the conductor, and body resistance (R) is measured with four surface electrodes placed on the right wrist and ankle. The skin was cleaned with 70% alcohol. Short- and long-term reproducibilities of resistance measurements produced coefficients of variation of 1.8 to 2.9%. 4,7 In our data, reproducibilities were r for measurements taken in the same subject within 1 wk (n 29) and r for repeat measurements up to 1 mo (n 40), or 2.5% variance. Dual-Energy X-ray Absorptiometry The DXA scanning technique measures the differential attenuation of two different levels of x-ray energy as they pass through the body, thereby allowing determination of bone-mineral content and soft-tissue mass on a pixel-by-pixel basis. The x-ray source (fan beam), mounted beneath the patient, generates a narrow, tightly collimated beam of x rays that pass through the patient at rapidly changing energies. The transmitted intensity of each energy level is measured by a radiation detector mounted on a movable arm directly above the x-ray source. The scanner uses an internal wheel to calibrate the bone-mineral component and an external lucite and aluminum phantom to determine the percentage of fat of each soft-tissue sample scanned. Simultaneous with the measurement of the skeleton, the percentage of fat is determined from the attenuation ratio of lower energy (70 kvp) to higher energy (140 kvp) detected by the beam. This ratio is calculated from all non-skeleton pixels scanned and extrapolated of the skeleton-containing pixels. The effective total body-radiation dose is 5.4 Sv. 20,21 The precisions of the measurements are 1.0% and 2.0% for FFM and FM, respectively, 22,23 which is consistent with or better than those reported in the literature. The FM derived from DXA measurements in previous studies correlated well with the FM determined by hydrodensitometry and total body 40K measurements. 23,24 All measurements were performed with a Hologic QDR-4500 instrument (Hologic, Waltham, MA, USA) and used Enhanced Whole Body 8.26a software. Statistics Descriptive statistics were calculated for height, weight, percentage of ideal body weight, BMI, and BIA parameters, including resistance, reactance, and H 2 /R and are expressed as mean standard deviation (SD). Simple regressions were calculated to test correlations of FFM between DXA and BIA. Student t test was used to test differences between methods. FFM as measured by DXA was used as the criterion measure-

66 250 Kyle et al. Nutrition Volume 17, Number 3, 2001 TABLE II. PREDICTION EQUATIONS FOR FFM USING ODD- AND EVEN-NUMBERED SUBJECTS Odd-numbered subjects (n 172) DXA FFM Predicted FFM Cross validation using even-numbered subjects FFM Even-numbered subjects (n 171) DXA FFM Predicted FFM Cross validation using odd-numbered subjects FFM Measured FFM kg (0.517 H 2 /R) (0.238 weight) (0.123 reactance) (4.077 sex: men 1, women 0) kg, r 0.987, SEE 1.75 kg kg, r 0.986, SEE 1.72 kg, TE 1.73 kg Measured FFM kg (0.520 H 2 /R) (0.221 weight) (0.140 reactance) (4.415 sex: men 1, women 0) kg, r 0.986, SEE 1.70 kg kg, r 0.987, SEE 1.76 kg, TE 1.78 kg DXA, dual-energy x-ray absorptiometry; FFM, fat-free mass; H 2 /R, height squared/resistance; r, validity coefficient; SEE, standard error of the estimate; TE, total error. ment. Stepwise multiple regression analysis was used to derive a prediction equation by BIA. Predictor variables, entered into the BIA model in the order of highest correlation coefficient and smallest standard error of estimation (SEE), were H 2 /R, weight, reactance, and sex. The prediction equation for BIA was developed by using a double cross-validation technique. 25 The subjects were assigned to two groups, odd and even, based on age (odd youngest, even second youngest). Thus, the two groups were evenly matched for age. An equation was developed for each, with the opposite group being used to cross validate each equation. If the equations proved to be similar (as evaluated by comparison of multiple r values and visible inspections of graphs), groups were combined and a single equation was developed using the entire sample. In addition to correlation and regression techniques, error analysis was performed. SEEs were calculated and used as errors of prediction for DXA-derived FFM by using BIA estimates. The technical error (TE) was calculated as TE n i 1 (FFM 1 FFM 2 ) 2 /n. 26 The total error of measurement estimates the magnitude of the error for a given measurement and is defined as the difference between measurements for the individual (i), i 1 n, where n is the number of individuals. 27 The total error is compared with the SEE. A large different between total error and SEE indicates poor accuracy of the prediction. Bland Altman analysis was calculated according to methods previously described 34 to assess the agreement between the two clinical measurements. The difference between the values was plotted against their means because the mean was the best available estimate of the true value. This analysis allows for the calculation of bias (estimated by the mean differences), the 95% confidence interval for the bias, and the limits of agreement (2 SDs of the difference). 28 Covariance analysis was used to compare the multiple regression models in the independent odd and even samples. Statistical significance was set at P 0.05 for all tests. RESULTS A total of 343 healthy adults aged y were recruited as subjects. Table I shows the anthropometric and BIA characteristics of healthy men and women grouped by age. Older subjects were progressively shorter than the younger ones. Weights increased with age in men and women until 59 y and thereafter decreased. Resistance decreased with age in both sexes until 59 y and then increased with age. Reactance decreased progressively in men and women after 60 y. H 2 /R also increased until age 59 y and then decreased in men and women. Age, height, weight, BMI, percentage of ideal body weight, resistance, reactance, and H 2 /R were not significantly different between the two groups (odd versus even; P 0.05, unpaired t test). Table II shows the BIA equation derived from the odd and even groups and the cross-validation results in the cross-validation sample. It is apparent from Table II that the validity coefficients (r values), SEE and TE, were similar between the even and odd samples and that the cross validation showed similar results. The regression lines (odd , and even ) were virtually identical, with deviation from the line of identities being similar for both samples. Thus, a single equation using all 343 subjects was developed for BIA prediction of FFM. The prediction equation developed from all subjects is shown in Table III. The order of entry of predictor variables was H 2 /R, weight, reactance, and sex for the BIA model (Table IV). H 2 /R accounted for 93% of the variability (SEE 2.8 kg) of the equation, whereas weight alone accounted for only 74% of the variability (SEE 5.5 kg) and height alone accounted for 75% of the variability (SEE 5.4 kg). Inclusion of height, weight, and age, without BIA parameters, accounted for 88% of the variability with a SEE of 3.7 kg. Thus, inclusion of BIA parameters clearly improved the prediction power and decreased the SEE compared with anthropometric parameters only. Replacing reactance with age in the multiple regressions decreased the correlation coefficient of the multiple regressions slightly (r ), increased SEE to 1.9 kg, and DXA FFM TABLE III. PREDICTION EQUATION FOR FFM USING ALL SUBJECTS* Predicted FFM Measured FFM kg (0.518 H 2 /R) (0.231 weight) (0.130 reactance) (4.229 sex; men 1, women 0) kg, r 0.986, SEE 1.72 kg, TE 1.74 kg * n 343 subjects. DXA, dual-energy x-ray absorptiometry; FFM, fat-free mass; H 2 /R, height squared/resistance; r, validity coefficient; SEE, standard error of the estimate; TE, total error.

67 Nutrition Volume 17, Number 3, 2001 Fat-Free Mass and Bioelectrical Impedance 251 TABLE IV. CONTRIBUTION AND ORDER OF ENTRY OF DEPENDENT VARIABLES TO THE BIA MODEL AND ANTHROPOMETRIC MODEL FOR FFM* Model and variables Cumulative dependent variables used in model Dependent variables r 2 SEE P r 2 SEE P BIA H 2 /R Weight Reactance Sex BMI Height Height weight Height weight age Height weight age sex * n 343 subjects. BIA, bioelectrical impedance analysis; BMI, body mass index; FFM, fat-free mass; H 2 /R, height squared/resistance; P, significance of contribution of each additional individual parameter to the stepwise multiple regression model; r 2, squared value of the validity coefficient; SEE, standard error of the estimate. resulted in an absolute mean differences of kg in 60-y-old men and kg women older than 80 y, thus slightly lowering the predictive accuracy of the multiple regressions. Table V shows the mean FFM and t test of the prediction equation odd, even, and combined by age group. Absolute mean difference in FFM between DXA and BIA equations ranged from kg to kg, and the t test was significant only in women older than 80 y. Covariance analysis was not significant (P 0.05) for odd and even groups for odd, even, and combined equations when adjusted for height, weight, age, and sex. Thus, the combined equation is valid to predict FFM in all age groups of men and women. Figure 1 shows the correlation and mean difference, according to Bland and Altmann, using the combined equation in all subjects. Subjects (data not shown) with BMIs above 27 kg/m 2 were analyzed separately to determine whether greater error occurred with the BIA equation in larger subjects. The mean difference for 67 subjects with BMIs above 27 was kg (P 0.05, paired t test; r 0.983, SEE 2.0 kg). Thus, it is possible to estimate FFM with the same equation for non-overweight and obese subjects (up to a BMI of 33.8 kg/m 2 ). DISCUSSION BIA has been developed for field use and has shown great potential for use in estimating body composition and it is easy, noninvasive, and inexpensive. BIA measures the conductivity of total body water and electrolytes that are found only in FFM with an applied radiofrequency electrical current. Both total body water and electrolytes have been shown to be highly correlated with FFM. 4,29 DXA is considered one of the reference methods for measuring FFM. 30 Our data provide measured (DXA) and predicted (BIA) FFM values for healthy subjects between ages of 22 and 94 y. Our study shows that a single BIA equation can be used to estimate FFM in healthy subjects with wide age spans and with BMIs ranging from 17 to 33.8 kg/m 2. Validity of BIA Heitmann 31 found that BIA had a significantly lower variability of estimates, making it the most accurate of the simpler methods (skinfold, BMI, and BIA). However, it is generally agreed that the accuracy of BIA depends on the variables included in the prediction equation and on using a specific prediction equation validated for a specific population. 17 Our equation for BIA-predicted FFM TABLE V. COMPARISON OF FAT-FREE MASS BY DXA AND BIA AS ESTIMATED BY EQUATIONS DETERMINED IN ODD, EVEN, AND ALL SUBJECTS Age (y) DXA Odd P* Even P* Combined P* Men Women * P 0.05, paired t test, DXA versus BIA. BIA, bioelectrical impedance analysis; DXA, dual-energy x-ray absorptiometry.

68 252 Kyle et al. Nutrition Volume 17, Number 3, 2001 FIG. 1. Correlations (top) and differences (bottom) of fat-free mass (FFM) in all subjects estimated by dual-energy X-ray absorptiometry (FFM DXA ) and bioelectrical impedance. The difference (calculated as FFM DXA FFM BIA per Bland Altman) is plotted against the mean of the measurements of FFM by DXA and BIA. SEE, standard error of the estimate; TE, technical error (see methods). Circles indicate men; triangles indicate women. used, in order of entry, H 2 /R, weight, reactance, and sex. Although H 2 /R alone accounted for 93% of the variability (Table IV), all other variables entered added significantly to FFM prediction by BIA. Furthermore, cross validation of specific equations is important to test for accuracy. Tables II and III show that all correlations were extremely high and similar for BIA (r 0.986) whether validation or cross-validation samples were used. Similarly, SEE and TE were low and consistent in all cases (1.76 kg). Therefore, the subject sample was combined and a single prediction equation was developed for all subjects. The correlation and mean differences of the combined sample are shown in Figure 1. Table V shows small, but non-significant differences for FFM in both men and women in all age groups except in women older than 80 y. Thus, BIA is able to predict FFM in men and women of various ages. Variations in BIA Parameters With Age Various BIA parameters have been shown to change with age (Table I) and quantity of FFM. In our study, mean resistance increased with age, suggesting a decrease in FFM. Resistance correlated significantly with weight (r 0.710, P ) and height (r 0.500, P ) but poorly with age (r 0.185, P ). In contrast, reactance decreased and correlated significantly with age (r 0.588, P ), moderately with weight (r 0.364, P ), but not with height (r 0.038, P 0.05). The ratios of reactance to height, reactance to weight, reactance to BMI, and reactance to FFM DXA decreased significantly (P ) with age. Both FFM and reactance decreased with age. Thus, reactance seems to be sensitive to the decrease in FFM with age and serves as an indicator of decreased FFM mass, decreased electric conductance with age, or both. Although this was not tested, the decrease in reactance might reflect sarcopenic or FFM wasting changes noted in older subjects. Further research is necessary to evaluate cause and effect of reactance on FFM. Although some investigators 1,8,10 have used age as a variable in BIA equations, the present results suggest that reactance is a more representative indicator of the decrease in FFM with age than the increase of resistance with age. Poor prediction of some BIA equations 1,6,9,11,12 may be related to the lack of inclusion of reactance in these equations and thus are valid for the age group for which they were developed but under- or overestimate FFM in younger or older subjects. In this study, replacing reactance with age in the multiple regressions analysis decreased the correlation coefficient of the multiple regressions slightly (r ), increased SEE to 1.9 kg, and resulted in absolute mean differences of kg in 60-y-old men to kg in women older than 80 y, thus slightly lowering the predictive accuracy of the multiple regressions. When applying existing BIA equations that included age to our subjects, we found poorer predictive ability in older subjects despite an age correction factor. For example, r 2 was 0.938, SEE was 2.7 kg, and absolute mean differences were kg in 20- to 29-y-old subjects and kg in subjects older than 80 y with the equation by Segal et al., 1 and r 2 was 0.89, SEE was 2.4 kg, and absolute mean differences were kg in 20- to 29-y-old subjects and kg in subjects older than 80 y with the equation by Deurenberg et al. 12 The equation by Segal et al. was developed in subjects y old and therefore would not compensate for changes in FFM in subjects older than 65 y. Recent results in our older subjects 32 suggested an accelerated loss of FFM in men and women older than 75 y. Therefore, it is possible that the curvilinear decrease in FFM with age was not adequately reflected by a linear age factor but was corrected better by reactance, which showed an accelerated decrease in older men and women. Reactance, therefore, appears to be essential for BIA equations developed for use in populations with large variations in age or that are intended for subjects where FFM is known to decrease with age, i.e., those older than 60 y. Hewitt et al. 33 also found that the addition of reactance significantly improved equations predicting FFM and total body water in older subjects. Theoretically, reactance is due to the capacitative effects of cell membranes and may be associated with changes in cell membrane permeability. Reactance has been reported to be sensitive to the distribution of total body weight between intra- and extracellular spaces. 8 Changes in intra- and extracellular fluid ratios reported in older versus younger subjects may be a reason that inclusion of reactance improved prediction of FFM in the present study. Comparisons of age-specific prediction equations for elderly subjects in our study showed lower correlations (Baumgartner et al., 34 r ; Roubenoff et al., 35 r ; Deurenberg et al., 11 r ), higher SEEs (Baumgartner et al., 2.8 kg; Roubenoff et al., 1.8 kg; Deurenberg et al., 2.4 kg), and larger absolute mean differences (Baumgartner et al., kg; Roubenoff et al., kg; Deurenberg et al., kg) than our equation. These comparisons suggest that there is no disadvantage or loss of predictive power by using the single BIA equation for all ages. Limitations of Study All subjects in this study were reported to be healthy. The BIA equation developed in this study is valid for healthy adults y

69 Nutrition Volume 17, Number 3, 2001 Fat-Free Mass and Bioelectrical Impedance 253 old. Slightly lower accuracy should be expected in women older than 80 y. Validity of the BIA equation in subjects with abnormal fluid balance is unknown and requires further validation. Further validation is also necessary in subjects with BMIs below 17.0 and above 33.8 kg/m 2. DXA was used as a criterion measurement to derive the prediction model in this study. The accuracy and precision of DXA depend in part on the thickness of the x-ray absorber. Beam hardening can occur for thick objects and insufficient attenuation of the x-ray beam can result in increased noise relative to the signal. The extent to which these factors may systematically affect the accuracy of estimates of lean soft tissue has not been established. 36 Another limitation of DXA is that it cannot be used to measure under- and overhydration. Although it accurately measures FFM, DXA does not distinguish between normally hydrated and over- or underhydrated FFM. DXA uses a constant of 73.2% for total body weight in FFM. Fuller et al. 37 found that the hydration fraction of FFM was % (range %), with no significant difference between men and women. Therefore, total body water should be determined simultaneously with the DXA for future validation studies of FFM. CONCLUSION The results of this study show that the new Geneva BIA equation, validated against DXA, can be used to predict FFM in subjects aged y and with BMIs from 17.0 to 33.8 kg/m 2. Furthermore, the results of this study suggest that reactance appears to be essential for BIA equations developed for use in populations with large variations in age and body mass or for subjects where FFM decrease with age, i.e., those older than 60 y. ACKNOWLEDGMENTS The authors are indebted to Giulio Conicella and to Luc Terraneo for technical assistance. The authors thank the Foundation Nutrition 2000Plus for its financial support. REFERENCES 1. Segal KR, Van Loan M, Fitzgerald PI, Hodgdon JA, Van Itallie TB. Lean body mass estimation by bioelectrical impedance analysis: a four-site cross over validation. Am J Clin Nutr 1988;47:7 2. Kyle UG, Pichard C. Dynamic assessment of fat-free mass during catabolism and recovery. Curr Opin Clin Nutr Metab Care 2000;3: Kushner RF, Schoeller DA. Estimation of total body water by bioelectrical impedance analysis. Am J Clin Nutr 1986;44: Lukaski HC. Assessment of fat-free mass using bioelectrical impedance measurements of the human body. Am J Clin Nutr 1985;41: Lukaski HC. Validation of tetrapolar bioelectrical impedance measurements to assess human body composition. J Appl Physiol 1986;60: Van Loan M, Mayclin P. Bioelectrical impedance analysis. Is it a reliable estimator of lean body mass and total body water? Hum Biol 1987;2: Jackson AS, Pollock ML, Graves JE, Mahar MT. Reliability and validity of bioelectrical impedance in determining body composition. J Appl Physiol 1988; 64: Lukaski HC, Bolonchuk WW. Estimation of body fluid volumes using tetrapolar bioelectrical measurements. Aviat Space Environ Med 1988;59: Graves JE, Pollack ML, Calvin AB, Van Loan M, Lohman TG. Comparison of different bioelectrical impedance analyzers in the prediction of body composition. Am J Hum Biol 1989;1: Heitmann BL. Prediction of body water and fat in adult danes from measurement of electrical impedance. A validation. Int J Obes 1990;14: Deurenberg P, van der Kooy K, Evers P, Hulshof T. Assessment of body composition by bioelectrical impedance in a population aged 60 y. Am J Clin Nutr 1990;51:3 12. Deurenberg P, van der Kooy K, Leenen R, Westrate JA, Seidell JC. Sex and age specific prediction formulas for estimating body composition from bioelectrical impedance: a cross-validation study. Int J Obes 1991;15: Roubenoff R, Baumgartner RN, Harris TB, et al. Application of bioelectrical impedance analysis to elderly population. J Gerontol 1997;52A:M Stolarczyk LM, Heyward VH, Van Loan MD, et al. The fatness-specific bioelectrical impedance analysis equations of Segal et al: are they generalizable and practical? Am J Clin Nutr 1997;66:8 15. Pichard C, Kyle UG, Slosman DO. Fat-free mass in chronic illness: comparison of bioelectrical impedance and dual-energy X-ray absorptiometry in 480 chronically ill and healthy subjects. Nutrition 1999;15: Pichard C, Kyle UG, Janssens JP, et al. Body composition by x-ray absorptiometry and bioelectrical impedance in chronic respiratory insufficiency patients. Nutrition 1997;13: Kyle U, Pichard C, Rochat T, et al. New bioelectrical impedance formula for patients with respiratory insufficiency: comparison to dual-energy X-ray absorptiometry. Eur Resp J 1998;12: Pichard C, Kyle U, Gremion G, Gerbase M, Slosman D. Body composition by X-ray absorptiometry and bioelectrical impedance in elite female runners. Med Sci Sports Exerc 1997;29: Hansen NJ, Lohman TG, Going SB, et al. Prediction of body composition in premenopausal females from dual-energy X-ray absorptiometry. J Appl Physiol 1993;75: Lewis MK, Blake GM, Fogelman I. Patient dose in dual x-ray absorptiometry. Osteoporos Int 1994;4: Blake GM, Patel R, Lewis MK. New generation dual x-ray absorptiometry scanners increase dose to patients and staff. J Bone Mineral Res 1996;11:S Mazess RB, Peppler WW, Gibbons M. Total body composition by dual-photon (153 Gd) absorptiometry. Am J Clin Nutr 1990;40: Slosman DO, Casez JP, Pichard C, et al. Assessment of whole-body composition using dual X-ray absorptiometry. Radiology 1992;185: Heymsfield SB, Wang J, Heshka S, Kehayias JJ, Pierson RN. Dual-photon absorptiometry: comparison of bone mineral and soft tissue mass measurements in vivo with established methods. Am J Clin Nutr 1989;49: Kerlinger FN, Pedhazur EJ. Multiple regression in behavioral research. New York: Holt, Rinehard & Winston, Wang ZM, Deurenberg P, Guo SS, et al. Six-compartment body composition model: inter-method comparisons of total body fat measurements. Int J Obes 1998;22: Wellens R, Chumlea WC, Guo S, et al. Body composition in white adults by dual energy X-ray absorptiometry, densitometry, and total body water. Eur J Clin Nutr 1993;59: Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;i: Kushner RF, Schoedler DA, Fjeld CR, Danford L. Is the impedance index (ht2/r) significant in predicting total body water? Am J Clin Nutr 1992;56: Mazess RB, Barden HS, Bisek JP, Hanson J. Dual-energy X-ray absorptiometry for total-body and regional bone-mineral and soft-tissue composition. Am J Clin Nutr 1990;51: Heitmann BL. Prediction of body water and fat in adult Danes from measurement of electrical impedance. A validation study. Int J Obes 1990;14: Kyle UG, Genton L, Karsegard L, et al. Cross-sectional changes in fat-free mass, skeletal muscle, body cell mass and fat mass between yrs. Eur J Clin Nutr (in press) 33. Hewitt MJ, Lohman TG, Going SB, Williams DP. Estimation of total body water by bioelectrical impedance analysis in older men and women. Med Sci Sports Exerc 1989;21:S Baumgartner RN, Heymsfield SB, Lichtman S, Wang J, Pierson RN Jr. Body composition in elderly people: effect of criterion estimates on predictive equations. Am J Clin Nutr 1991;53: Roubenoff R. Applications of bioelectrical impedance analysis for body composition to epidemiologic studies. Am J Clin Nutr 1996;64:459S 36. Laskey MA, Lyttle KD, Flaxman ME, Barber RW. The influence of tissue depth and composition on the performance of the Lunar dual-energy X-ray absorptiometer whole-body scanning mode. Eur J Clin Nutr 1992;46: Fuller NJ, Jebb SA, Laskey MA, Coward WA, Elia M. Four-component model for the assessment of body composition in humans: comparison with alternative methods, and evaluation of the density and hydration of fat-free mass. Clin Sci 1992;82:687

70 APPENDIX 4

71 Clinical Section Gerontology 2001;47: Received: January 15, 2001 Accepted: March 22, 2001 Comparison of Four Bioelectrical Impedance Analysis Formulas in Healthy Elderly Subjects Laurence Genton a Véronique L. Karsegard a Ursula G. Kyle a Didier B. Hans b Jean-Pierre Michel c Claude Pichard a a Clinical Nutrition, b Nuclear Medicine and c Geriatrics, Geneva University Hospital, Geneva, Switzerland Key Words Body composition W Bioelectrical impedance Abstract Background: Changes of body composition occur with aging and influence health status. Thus accurate methods for measuring fat-free mass (FFM) in the elderly are essential. Objective: The purpose of this study was to compare FFM obtained by three bioelectrical impedance analysis (BIA) published formulas specific for the elderly and one equation intended for all age groups, with FFM derived from dual-energy X-ray absorptiometry (FFM DXA ), in healthy elderly subjects. Methods: Healthy Caucasian subjects over 65 years (106 women, age 75 B 6.2, body mass index 25.2 B 4.1 and 100 men, age 74.6 B 6.6, body mass index 25.8 B 3.0) were measured by DXA (Hologic QDR-4500) and BIA (Xitron, 50 khz). FFM BIA was calculated by the published formulas of Deurenberg, Baumgartner, Roubenoff and Kyle and compared to FFM DXA by a Bland-Altman analysis. Results: The Deurenberg and Roubenoff BIA formulas underestimated FFM compared to DXA by 7.1 and 2.9 kg in women and 6.7 and 2.3 kg in men, respectively. The Baumgartner formula overestimated FFM by 4.3 kg in women and 1.4 kg in men. The Kyle formula showed differences of 0.0 kg in women and 0.2 kg in men, and the limits of agreement of FFM BIA (Kyle) relative to FFM DXA were 3.3 and +3.3 kg for women and 3.8 and +4.3 kg for men. Conclusion: The Kyle BIA formula accurately predicts FFM in elderly Swiss subjects between 65 and 94 years, with a body mass index of 17 to 34.9 kg/m 2. The other BIA formulas developed especially for the elderly are not valid in this population. Introduction Copyright 2001 S. Karger AG, Basel Significant changes in body composition occur with aging [1 3] and influence overall health. Increases of total and abdominal fat masses (FM) have been associated with cardiovascular and metabolic diseases [4, 5], while the decline of fat-free mass (FFM) may impair muscle strength, immunity, and increase morbidity and mortality [6, 7]. Prevention and treatment of these body composition-related diseases require accurate and precise methods for measuring FM and FFM. Dual-energy X-ray absorptiometry (DXA), hydrodensitometry, whole body counting and isotope dilution are considered reference methods, but they are expensive, require extensive operator training and cannot be per- ABC Fax karger@karger.ch S. Karger AG, Basel X/01/ $17.50/0 Accessible online at: Claude Pichard, MD, PhD Clinical Nutrition and Diet Therapy, Geneva University Hospital CH 1211 Geneva (Switzerland) Tel , Fax , pichard@cmu.unige.ch

72 formed at the patient s bedside. Their use for clinical practice and studies is therefore limited. In large studies, body composition is frequently evaluated by body mass index (BMI; weight/height 2 ), skinfold thickness or bioelectrical impedance analysis (BIA). However, the ability of BMI to estimate FM is much debated [8, 9]. In women, a high BMI is associated with high FM and FFM, whereas in men, it may reflect only high FM [8]. Skinfold thickness may lead to errors in body composition prediction in the elderly because it does not measure the increase of intra-abdominal FM occurring with age [10]. Considering the limitations of these two methods, research is presently focusing on BIA. BIA is an easy, quick, noninvasive and operator-independent method for measuring body composition, which has been validated against DXA in healthy [11] and ill adults [12]. Recently, reference values of FM and FFM measured by BIA have been published for healthy adults aged years [13]. However, its validity in the elderly remains to be demonstrated. To our knowledge, only three BIA formulas for the elderly have been published to date: by Deurenberg et al. [14], Baumgartner et al. [15] and Roubenoff et al. [16], in selected populations of the Netherlands, New York, USA and Massachusetts, USA, respectively. These formulas may not be valid in the elderly Swiss population, since BIA formulas are often population-specific [17]. On the other hand, Kyle et al. [18] developed a BIA formula specific for the Swiss population, for all ages, but only 42 men and 48 women over 70 years were included. The purpose of this study was to evaluate FFM obtained by the three previously mentioned BIA equations developed specifically for the elderly and the non-age-specific equation of Kyle et al., with FFM derived by DXA (FFM DXA ), in a Caucasian population over 65 years. This comparison aimed at determining the best predictive BIA equations in the elderly population in the Geneva area. A valid equation would allow to establish reference values of FM and FFM in people over 65 years old and, consequently, help to detect and treat malnutrition in elderly subjects. Methods Subjects Healthy ambulatory subjects over 65 years were recruited through advertisements in newspapers and clubs for elderly people, in the Geneva area. During the enrollment telephone call, they were screened for acute and chronic illness. People with decompensated cardiac, pulmonary, renal or hepatic failure, symptomatic neurological disorders (Parkinson s disease, plegia, paresis), known active cancer or infectious diseases, significant mental impairment, involuntary weight loss or gain (1 3% total body weight in the last 6 months), and hospitalization in the last 6 months were excluded. We included the first 109 women and 105 men who were eligible for the study. They presented themselves at the University Hospital of Geneva, between 7.30 and 9.00 a.m. for one 2-hour session, after having consumed 1.5 liters of fluid the day before to avoid dehydration. They first answered a detailed questionnaire about their medical history, presently taken drugs, professional training and physical activity. Then, they underwent body composition measurements by BIA and DXA. All subjects signed an informed consent statement and the study protocol was approved by the Ethical Committee of the Department of Medicine, Geneva University Hospital. Anthropometric Measurements Body height was measured to the nearest 0.5 cm with a height gauge and body weight to the nearest 0.1 kg on a balance beam scale (Seca, Germany), with the subject in underclothing without shoes. BMI was calculated as weight (kg)/height 2 (m). Thigh circumference was measured below the gluteal fold on the nondominant side, with a plastic tape, to the nearest 0.5 cm. Bioelectrical Impedance Analysis In theory, the estimation of body composition by BIA relies on the geometrical relationship between impedance (Z), length (L) and volume (V) of an electrical conductor: V = ÚL 2 /Z (Ú is the specific resistivity, in ohms). Adapted to the human body, V corresponds to the volume of FFM, and L to the height of the subject. Z is composed of the pure resistance R of the conductor, the FFM, and the reactance Xc produced by the capacitance of cellular membranes, tissue interfaces and nonionic tissues: Z 2 = R 2 + Xc 2. Practically, the skin was cleaned with 70% alcohol. Adhesive electrodes (3M Red Dot, *M Health Care, Borken, Germany) were placed on the right hand and foot, the subject lying on his back as previously described [19]. A generator (Xitron, 4000B, Xitron Technologies, San Diego, Calif., USA) applied an alternating electrical current of 50 khz and 0.8 ma to these electrodes. The measured resistance and reactance were used to calculate FFM by the following formulas: Deurenberg: FFM = ht 2 /R sex [14] Baumgartner: FFM = ht 2 /R wt sex leg circumference [15] Roubenoff: FFM (women) = ht 2 /R wt reactance [16] FFM (men) = ht 2 /R wt reactance [16] Geneva: FFM = ht 2 /R wt reactance sex [18], where ht = height, R = resistance, wt = weight and sex is coded 0 for women and 1 for men. Short-term and long-term precision of FFM measured by BIA indicates coefficients of variation of % in healthy adults [20, 21]. Error of prediction for percent FM BIA is 2.7% in subjects aged years compared to densitometry [22]. Dual-Energy X-Ray Absorptiometry Body composition was determined with DXA, Hologic QDR 4500 (Hologic Inc., Waltham, Mass., USA), enhanced 8.26 wholebody software version. In this scanning technique, an X-ray generator emits switched pulsed radiation of two energies, 100 and 140 kvp, in a fan-beam mode. As they pass through the body, these two X-ray beams are attenuated due to the absorption and scattering of 316 Gerontology 2001;47: Genton/Karsegard/Kyle/Hans/Michel/ Pichard

73 Table 1. Anthropometric and BIA characteristics of the study subjects Women (n = 106) mean B SD range Men (n = 100) mean B SD range Age, years 75.0B B Height, cm 159.7B B Weight, kg 64.4B B BMI, kg/m B B Thigh circumference, cm 55.9B B Resistance, ø 571.3B B Reactance, ø 55.0B B Phase angle, degrees 5.5B B Height 2 /resistance, cm 2 /ø 45.2B B the photons. The attenuation is measured for every pixel of the body surface by a linear array of 216 detectors. A complex development measurement allows determination of bone mineral and soft tissue densities. Soft tissue can further be partitioned into FM and FFM, since they have different attenuation characteristics. To our knowledge, no information on precision of QDR 4500 for measuring body composition is available, but the precision of DXA QDR 2000, a former model of Hologic, is 1.0% for FFM and 2.0% for FM [23]. The FM derived from DXA measurements in previous studies correlates well with the FM determined by hydrodensitometry and total body K 40 measurements [23, 24]. Statistical Analysis Statistical analysis was performed with Statview 5.0 (Abacus Concept, Berkeley, Calif., USA). Results are expressed as mean B standard deviation. Simple regressions were calculated to test correlations between FFM measured by DXA and the BIA formulas. The technical error of measurement was calculated as (FFM BIA FFM DXA ) 2 /2 n. It represents the dispersion of the differences and for a normal distribution, 95% of the values for a measurement should be within plus or minus twice the technical error value of the corresponding true value. Ideally, body composition methods should have low TE and high correlation coefficients [25]. Agreement between DXA and each single BIA formula was assessed by a Bland-Altman analysis [26]: the difference between FFM BIA and FFM DXA is plotted against the mean of FFM BIA and FFM DXA, the mean being the best available estimate of the true value. Standard erros and limits of agreement (mean difference B two standard deviations) were calculated. This analysis shows whether FFM is under- or overestimated by one BIA formula, whether the scatter of differences increases or decreases over the range of FFM and whether the intermethod differences become more negative or positive over the range of FFM while the scatter remains the same. Results One hundred and six women and 100 men were considered for analysis. Table 1 shows their anthropometric and BIA characteristics. We excluded individuals who were not Caucasian (2 women, 4 men) or who did not have either a DXA or BIA measurement (1 woman, 1 man). Four women had a BMI below 18.5 kg/m 2 but no apparent health problem and were therefore included for analysis. FFM BIA (kg), calculated by the formulas of Deurenberg, Baumgartner, Roubenoff and Kyle, and FFM DXA (kg) are presented in table 2, as well as the mean differences between FFM obtained by each of the BIA formulas and FFM DXA. Compared to DXA, FFM BIA (kg) was underestimated by Deurenberg and Roubenoff, and overestimated by Baumgartner, although only slightly in men. On the other hand, FFM BIA (Kyle) related well to FFM DXA with mean differences of 0.0 kg in women (SD = 1.6 kg) and 0.2 kg in men (SD = 2.0 kg). FFM measured by all BIA formulas correlated well with FFM DXA (r between 0.87 and 0.94), but the regression slopes showed large variations. FFM measured by the formula of Kyle showed the slope closest to 1 (women 0.98, men 0.95). The Bland-Altman plots confirmed the tendencies shown by the descriptive statistics and the simple linear regression analysis (fig. 1, 2). The FFM obtained by the Kyle formula showed the best agreement with the FFM DXA. The limits of agreement were 3.25 and kg for women and 3.77 and kg for men, meaning that 95% of differences will lie between these limits, which were very good compared to the other BIA formulas. Furthermore, the measurements were symmetrically distributed along the line of equality, independently of the mean FFM, suggesting that the errors were not different between individuals with low and high FFM. Bioelectrical Impedance Analysis Formulas in the Elderly Gerontology 2001;47:

74 Table 2. FFM derived from DXA and BIA Mean B SD kg FFM B SD kg r Slope B S E S EE TE Women (n = 106) DXA 41.3B4.6 BIA Deurenberg et al. [14] 34.2B B B Baumgartner et al. [15] 45.6B B B Roubenoff et al. [16] 38.4B B B Kyle et al. [18] 41.3B B B Men (n = 100) DXA 55.9B5.9 BIA Deurenberg et al. [14] 49.2B B B Baumgartner et al. [15] 57.3B B B Roubenoff et al. [16] 53.6B B B Kyle et al. [18] 56.1B B B FFM = FFM BIA FFM DXA; r = correlation coefficient; SE = standard error; SEE = standard error of the estimate; TE = technical error of measurement = (FFM BIA FFM DXA ) 2 /2 n. Discussion In this study, FFM calculated by the three BIA formulas developed especially for elderly differed significantly from FFM DXA. These discrepancies are difficult to explain since the population selection and characteristics, the reference methods for measuring body composition and the parameters used in the BIA formulas vary from one study to the next. However, it appears that the validity of BIA formulas in the elderly is population-specific. On the other hand, the Kyle formula compared very well with DXA [18]. Population Selection and Characteristics The Deurenberg and Kyle formulas were developed in European populations, while the two others were developed in white American populations [14 16, 19]. Thus, we intuitively thought that the European formulas would compare better with DXA in our population than the American ones. This assumption is true only for the Kyle formula, developed in the Geneva area but based on a study population with a wide age range (20 94 years) [19]. Thus, it seems that the validity of a BIA formula depends more on ethnicity than on age. Mean age and BMI of the reference populations of Deurenberg et al. [14], Baumgartner et al. [15] and Roubenoff et al. [16] differed only slightly from ours. To evaluate the influence of these factors on FFM in kilograms, we classified our subjects into age groups (65 74, 174 years) or BMI groups (!25, 25 30, 130 kg/m 2 ) and compared FFM BIA, obtained by each BIA formula, and FFM DXA by using a Bland-Altman test (table 3). The mean differences between FFM BIA and FFM DXA, and the 95% confidence interval remained similar between age groups for all BIA formulas, whereas these parameters became larger as the BMI increased, except when using the Baumgartner formula in men. However, in all BMI groups, FFM BIA (Kyle) shows the least difference with FFM DXA (!18.5 kg/m 2 : 0.2 kg, : +0.2 kg, 125: +2.1 kg) and the 95% confidence interval remained similar even if the limits of agreement increased. Reference Methods and Their Limits DXA was the reference method used in our study, as well as in the studies by Roubenoff et al. [16] and Kyle et al. [19]. One limitation of this method is large subject sizes, which may lead to an overestimation of percent FFM [27]. However, we had no individuals with a BMI over 35 in our study and therefore errors due to this limitation should be minimal. The second limitation is that the measured FFM includes not only appendicular skeletal muscle but also nonmuscle components such as skin, neurovascular tissues, connective tissue, interstitial fat and extracellular fluid. Some authors describe an increase 318 Gerontology 2001;47: Genton/Karsegard/Kyle/Hans/Michel/ Pichard

75 Fig. 1. Bland-Altman plots for women, comparing FFM DXA and FFM BIA obtained by each BIA formula. Plain lines show mean (FFM BIA FFM DXA ), (FFM BIA FFM DXA ) + 2 SD and (FFM BIA FFM DXA ) 2 SD. The dotted line represents the line of equality between the methods, i.e. the line on which the measurements should lie if the two methods were equal. with age of these last three elements, which may result in an overestimation of FFM in the elderly compared to other reference methods [3, 28]. Furthermore, soft-tissue algorithms rely on a fixed hydration of FFM (0.73 ml/g) [29], which may be questionable in the elderly. Nevertheless, Visser et al. [30] validated DXA in the elderly against a four-compartment model. The described limitations of DXA are the same for all DXA instruments, and we cannot explain why FFM BIA (Roubenoff), validated against DXA, shows large dif- Bioelectrical Impedance Analysis Formulas in the Elderly Gerontology 2001;47:

76 Fig. 2. Bland-Altman plots for men, comparing FFM DXA and FFM BIA obtained by each BIA formula. Plain lines show mean (FFM BIA FFM DXA ), (FFM BIA FFM DXA ) + 2 SD and (FFM BIA FFM DXA ) 2 SD. The dotted line represents the line of equality between the methods, i.e. the line on which the measurements should lie if the two methods were equal. ferences with DXA in our study population. One explanation could be the use of different DXA instruments, but the Lunar instrument, used by Roubenoff (Lunar DPX- L), underestimates percent FM ( 3.4%), thus overestimates percent FFM, compared to the Hologic instrument, used in our study [31]. We would therefore expect a higher FFM BIA (Roubenoff) than FFM DXA in our study, which is not the case. Hydrodensitometry, used by Deurenberg et al. [14], assumes a constant density of FM (0.9 g/ml) and FFM 320 Gerontology 2001;47: Genton/Karsegard/Kyle/Hans/Michel/ Pichard

77 Table 3. Differences of FFM calculated as FFM measured by BIA minus DXA, and their limits of agreement, by age and BMI groups n Deurenberg et al. [14] Baumgartner et al. [15] Roubenoff et al. [16] Kyle et al. [18] FFM agreement FFM agreement FFM agreement FFM agreement Women All , , , , 3.3 Age, years , , , , , , , , 2.8 BMI! , , , , , , , , kg/m , , , , 5.0 Men All , , , , 4.1 Age, years , , , , , , , , 3.9 BMI! , , , , , , , , kg/m , , , , 6.0 FFM = FFM BIA FFM DXA; agreement = limits of agreement. (1.1 g/ml) [32]. Some studies confirm the constancy of FFM density in the elderly [33, 34]. Goran et al. [33] compared body composition measured by hydrodensitometry and a four-compartment model in 82 people aged years. He found that, despite a calculated FFM density of 1.1 g/ml in his subjects, the Siri equation, deducing percent FM from body density, is not valid in elderly women. Other studies describe a lower density of FFM due to a decrease in bone mineralization, total body protein and an increase in total body water [32, 35], and mention that the resulting underestimation of FFM should be of 1 2 kg [14, 15]. Considering that two-compartment models do not account for the heterogeneity of FFM, resulting in potential errors, measuring directly one or several components of FFM may be preferable. Baumgartner et al. [15] used a four-compartment model based on hydrodensitometry, 3 H 2 O dilution and DXA (Lunar DP-4) to develop their BIA formula. Although this method is supposedly more accurate for measuring FFM, one must keep in mind that each separate measurement leads to its own errors. A study including 18 elderly women showed an underestimation of 5.3 B 3.8% FM by DXA relative to a fourcompartment model [35]. The underestimation of FM by DXA was confirmed by Wang et al. [25], who used a sixcompartment model. Unfortunately, determination of body composition with a multicompartment model necessitates sophisticated equipment, is time-consuming and cannot be performed easily in the elderly. Comparison of BIA Formulas The BIA formula of Deurenberg clearly underestimates FFM in our study population, compared to DXA. This may be explained by an underestimation of FFM by their reference method, hydrodensitometry, by a potential overestimation of FFM by DXA, as previously described, or by the BIA formula itself. They recognized that the use of other BIA formulas led to a 6- to 7-kg increase of FFM. The overestimation of percent FM by this formula, thus underestimation of FFM in the elderly is confirmed by other studies [29, 36]. Their formula does not include weight, but this element seems less important than other BIA parameters [37]. The underestimation of FFM by the formula of Roubenoff was less than by Deurenberg. It may be explained by a lower ratio height 2 /resistance in their study than ours. Their subjects were considerably shorter ( 3 cm for women, 3.4 cm for men), while the resistance was only slight- Bioelectrical Impedance Analysis Formulas in the Elderly Gerontology 2001;47:

78 ly decreased. Roubenoff confirms that the new equation overestimated FFM when applied to taller and less obese subjects. Furthermore, in his study, DXA and BIA measurements were not performed on the same day but within 2 weeks, leading to a more difficult interpretation of intermethod FFM differences. The importance of reactance used in their formula is unclear. The formula of Baumgartner is not accurate in our elderly women, who had larger upper thigh circumferences (55.9 B 5.8 cm) than his reference population (47.5 B 4.4 cm). This difference may result from a less important compression of FM during our measurements of thigh circumference, or from a different FM and FFM distribution in our population. Future research on segmental body composition may clarify this issue. Additionally, it is not quite clear why thigh and not arm circumference was included. Is thigh circumference supposed to reflect FFM? However, the resistance measured by BIA is mainly composed of the resistance of the extremities, including arms [38]. Limitations of BIA in Elderly BIA predicts total body water and deduces FFM by assuming a constant hydration of FFM. However, it is still unclear how aging influences total hydration and body fluid distribution and therefore BIA measurements. In a study including 200 black and white women, percentage of total hydration decreased by 3.7% over the age span of years, while the ratio of total body water/ffm increased [32]. This result is not consistent with Mazariegos et al. [39] who found similar hydration of FFM between young and elderly women of Guatemala. Thus, accuracy of BIA in the elderly would rely on accurate determination of total body water and its intra- and extracellular compartments. Furthermore, total body resistance measured by BIA depends mainly on the resistance of the extremities. Since adipose tissue thickness decreases with age in arms and legs, and intra-abdominal and subcutaneous adipose tissue increases in the trunk [3, 40, 41], it would appear that the formula of Kyle, developed mainly in younger adults, could not measure FFM appropriately in the elderly. It is unclear why this assumption is not verified in this study and why one single formula can be used for all ages. Conclusion The Kyle BIA formula, based mainly on young and middle-aged subjects, predicts FFM best in the healthy Swiss elderly, between 65 and 94 years and with BMI of kg/m 2. Thus, one single BIA formula can be used for men and women of all ages. This finding facilitates interpretations of body composition measurements in clinical follow-up and allows for easy comparison of subjects with various ages and BMI. Acknowledgements We thank Mrs. Bernadette Mermillod, PhD, Division of Medical Informatics, for her help in statistics, Prof. René Rizzoli for his full editorial comments, the technicians of Nuclear Medicine for doing the DXA scans, and the Foundation Nutrition 2000 Plus for financial support. References 1 Flynn MA, Nolph GB, Baker AS, Krause G: Aging in humans: A continuous 20-year study of physiologic and dietary parameters. J Am Coll Nutr 1992;11: Oldroyd B, Stewart SP, Truscott JG, Westmacott CF, Smith MA: Age related changes in body composition. Appl Radiat Isot 1998;49: Baumgartner RN, Stauber PM, McHugh D, Koehler KM, Garry PJ: Cross-sectional age differences in body composition in persons 60+ years of age. J Gerontol 1995;50A:M307 M Poehlman ET, Gardner AW, Goran MI, Arciero PJ, Toth MJ, Ades PA, Calles-Escadon J: Sympathetic nervous system activity, body fatness, and body fat distribution in younger and older males. J Appl Physiol 1995;78: Bertrais S, Balkau B, Vol S, Forhan A, Calvet C, Marre M, Eschwege E: Relationships between abdominal body fat distribution and cardiovascular risk factors: An explanation for women s healthier cardiovascular risk profile. The D.E.S.I.R. Study. Int J Obes Relat Metab Disord 1999;23: Heitmann BL, Erikson H, Ellsinger BM, Mikkelsen KL, Larsson B: Mortality associated with body fat, fat-free mass and body mass index among 60-year-old Swedish men A 22- year follow-up. The study of men born in Int J Obes Relat Metab Disord 2000;24: Karlsson M, Nilsson JA, Sembo I, Redlund- Johnell I, Johnell O, Obrant KJ: Changes of bone mineral mass and soft tissue composition after hip fracture. Bone 1996;18: Morabia A, Ross A, Curtin F, Slosman DO, Pichard C: Relation of body mass index to a dual-energy X-ray absorptiometry measure of fatness. Br J Nutr 1999;82: Roubenoff R, Dallal GE, Wilson PW: Predicting body fatness: The body mass index vs estimation by bioelectrical impedance. Am J Public Health 1995;85: Han TS, Carter R, Currall JE, Lean ME: The influence of fat free mass on prediction of densitometric body composition by bioelectrical impedance analysis and anthropometry. Eur J Clin Nutr 1996;50: Pichard C, Kyle UG, Gremion G, Gerbase M, Slosman DO: Body composition by X-ray absorptiometry and bioelectrical impedance in female runners. Med Sci Sports Exerc 1997;29: Gerontology 2001;47: Genton/Karsegard/Kyle/Hans/Michel/ Pichard

79 12 Pichard C, Kyle UG, Janssens JP, Burdet L, Rochat T, Slosman DO, Fitting JW, Thiebaud D, Roulet M, Tschopp JM, Landry M, Schutz Y: Body composition by X-ray absorptiometry and bioelectrical impedance in chronic respiratory insufficiency patients. Nutrition 1997;13: Pichard C, Kyle UG, Bracco D, Slosman DO, Morabia A, Schutz Y: Reference values of fatfree and fat mass by bioelectrical impedance analysis (BIA) in 3393 healthy subjects. Nutrition 2000;16: Deurenberg P, van der Kooji K, Evers P, Hulshof T: Assessment of body composition by bioelectrical impedance in a population aged 1 60 years. Am J Clin Nutr 1990;51: Baumgartner RN, Heymsfield SB, Lichtman S, Wang J, Pierson RN Jr: Body composition in elderly people: Effect of criterion estimates on predictive equations. Am J Clin Nutr 1991;53: Roubenoff R, Baumgartner RN, Harris TB, Dallal GE, Hannan MT, Economos CD, Stauber PM, Wilson PW, Kiel DP: Application of bioelectrical impedance analysis to elderly population. J Gerontol 1997;52A:M129 M Bussolotto M, Ceccon A, Sergi G, Giantin V, Beninca P, Enzi P: Assessment of body composition in elderly: Accuracy of bioelectrical impedance analysis. Gerontology 1999;45: Kyle UG, Genton L, Karsegard L, Slosman DO, Pichard C: Single prediction equation for bioelectrical impedance analysis in adults aged years. Nutrition, in press. 19 Kushner RF, Schoeller DA: Estimation of total body water by bioelectrical impedance analysis. Am J Clin Nutr 1986;44: Lukashi HC, Johnson PE, Bolonchuk WW, Lykken GI: Assessment of fat-free mass using bioelectrical impedance measurements of the human body. Am J Clin Nutr 1985;41: Jackson AS, Pollock ML, Graves JE, Mahar MT: Reliability and validity of bioelectrical impedance in determining body composition. J Appl Physiol 1988;64: Lukaski HC, Bolonchuk WW, Hall CB, Siders WA: Validation of tetrapolar bioelectrical impedance method to assess human body composition. J Appl Physiol 1986;60: Slosman DO, Casez JP, Pichard C, Rochat T, Fery F, Rizzoli R, Bonjour JP, Morabia A, Donath A: Assessment of whole-body composition using dual X-ray absorptiometry. Radiology 1992;185: Heymsfield SB, Wang J, Heshka S, Kehayias JJ, Pierson RN: Dual-photon absorptiometry: Comparison of bone mineral and soft tissue mass measurements in vivo with established methods. Am J Clin Nutr 1989;49: Wang ZM, Deurenberg P, Guo SS, Pietrobelli A, Wang J, Pierson RN, Heymsfield SB: Sixcompartment body composition model: Intermethod comparisons of total body fat measurement. Int J Obes Relat Metab Disord 1998;22: Bland JM, Altman DG: Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;i: Formica CA: Total body bone mineral and body composition by absorptiometry; in Blake GM, Wahner HW, Fogelman I (eds): The Evaluation of Osteoporosis: Dual Energy X-Ray Absorptiometry and Ultrasound in Clinical Practice. London, Dunitz, 1999, pp Heymsfield SB, Smith R, Aulet M, Benson B, Lichtman S, Wang J, Pierson RN Jr: Appendicular skeletal muscle mass: Measurement by dual-photon absorptiometry. Am J Clin Nutr 1990;52: Ravaglia G, Fort P, Maioli F, Boschi F, Cicognani A, Gasbarrini G: Measurement of body fat in healthy elderly men: A comparison of methods. J Gerontol 1999;54A:M70 M Visser M, Fuerst T, Lang T, Salamone L, Harris T: Validity of fan-beam dual-energy X-ray absorptiometry for measuring fat-free mass and leg muscle mass. Health, Aging, and Body Composition Study Dual-Energy X-ray Absorptiometry and Body Composition Working Group. J Appl Physiol 1999;87: Tothill P, Avenell A, Love J, Reid DM: Comparisons between Hologic, Lunar and Norland dual-energy X-ray absorptiometers and other techniques used for whole-body soft tissue measurements. Eur J Clin Nutr 1994;48: Aloia JF, Vaswani A, Flaster E, Ma R: Relationship of body water compartments to age, race, and fat-free mass. J Lab Clin Invest 1998; 132: Goran MI, Toth MJ, Poehlman ET: Assessment of research-based body composition techniques in healthy elderly men and women using the 4-compartment model as a criterion method. Int J Obes Relat Metab Disord 1998;22: Mazariegos M, Wang ZM, Gallagher D, Baumgartner RN, Allison DB, Wang J, Pierson RN Jr, Heymsfield SB: Differences between young and old females in the five levels of body composition and their relevance to the two-compartment chemical model. J Gerontol 1994;49: M201 M Bergsma-Kadijk JA, Baumeister B, Deurenberg P: Measurement of body fat in young and elderly women: Comparison between a fourcompartment model and widely used reference methods. Br J Nutr 1996;75: Reilly JJ, Murray LA, Wilson J, Durnin JV: Measuring the body composition of elderly subjects: A comparison of methods. Br J Nutr 1994;72: Kotler DP, Burastero S, Wang J, Pierson RN Jr: Prediction of body cell mass, fat-free mass, and total body water with bioelectrical impedance analysis: Effects of race, sex, and disease. Am J Clin Nutr 1996;64:489S 497S. 38 NIH Technology Assessment Conference Statement: Bioelectrical Impedance Analysis in Body Composition Measurement. 1994, pp Mazariegos M, Valdez C, Kraaij S, Van Setten C, Liurink C, Breuer K, Haskell M, Mendoza I, Solomons NW, Deurenberg P: A comparison of body fat estimates using anthropometry and bioelectrical impedance analysis with distinct prediction equations in elderly persons in the Republic of Guatemala. Nutrition 1996;12: Chumlea WC, Guo SS, Kuczmarski RJ, Vellas B: Bioelectrical and anthropometric assessments and reference data in the elderly. J Nutr 1993;123: Gallagher D, Visser M, De Meersman RE, Sepulveda D, Baumgartner RN, Pierson RN, Harris T, Heymsfield SB: Appendicular skeletal muscle mass: Effects of age, gender, and ethnicity. J Appl Physiol 1997;83: Bioelectrical Impedance Analysis Formulas in the Elderly Gerontology 2001;47:

80 APPENDIX 5

81 APPLIED NUTRITIONAL INVESTIGATION Fat-Free and Fat Mass Percentiles in 5225 Healthy Subjects Aged 15 to 98 Years Ursula G. Kyle, MS, RD, Laurence Genton, MD, Daniel O. Slosman, MD, and Claude Pichard, MD, PhD From the Department of Clinical Nutrition and the Department of Nuclear Medicine, Geneva University Hospital, Geneva, Switzerland OBJECTIVES: Fat-free mass (FFM) and fat mass (FM) are important in the evaluation of nutritional status. Bioelectrical impedance analysis (BIA) is a simple, reproducible method used to determine FFM and FM. Because normal values for FFM and FM have not yet been established in adults aged 15 to 98 y, its use is limited in the evaluation of nutritional status. The aims of this study were to determine reference values for FFM, FM, and percentage of FM by BIA in a white population of healthy adults, observe their differences with age, and develop percentile distributions for these parameters between ages 15 and 98 y. METHODS: Whole-body resistance and reactance of 2735 healthy white men and 2490 healthy white women, aged 15 to 98 y, was determined by 50-kHz BIA, with four skin electrodes on the right hand and foot. FFM and FM were calculated by a previously validated, single BIA formula and analyzed for age decades. RESULTS: Mean FFM peaked in 35- to 44-y-old men and 45- to 54-y-old women and declined thereafter. Mean FFM was 8.9 kg or 14.8% lower in men older than 85 y than in men 35 to 44 y old and 6.2 kg or 14.3% lower in women older than 85 y than in women 45 to 54 y old. Mean FM and percentage of FM increased progressively in men and women between ages 15 and 98 y. The results suggested that the greater weight noted in older subjects is due to larger FM. CONCLUSIONS: The percentile data presented serve as reference to evaluate deviations from normal values of FFM and FM in healthy adult men and women at a given age. Nutrition 2001;17: Elsevier Science Inc KEY WORDS: bioelectrical impedance analysis, fat-free mass, body fat, fat-free mass measured with bioelectrical impedance analysis, body composition, reference standard, sex INTRODUCTION Significant changes in body composition occur with aging and are believed to be a consequence of imbalances between energy intake and energy needs associated with an increasingly sedentary lifestyle. Progressive increases in fat mass (FM) and progressive reductions in fat-free mass (FFM) have been noted. In adults, overand undernutrition contribute to increased mortality and morbidity. In the elderly, the age-related loss of muscle mass, or sarcopenia, is prevalent and strongly associated with impaired mobility, increased morbidity and mortality, and lower quality of life. 1 Much of our current understanding of the changes in body composition with advancing age comes from studies that are 30 y old. 2 National surveys for reference data for body composition measures that have large samples and are suitable for describing differences between individuals and between levels of muscle and FM are needed. Because weight and body mass index (BMI) alone are not an adequate guide of underlying changes in FFM and FM during menopause 3 and aging in general, 4 body composition should be measured in clinical management programs and epidemiological and clinical studies of aging. 4 Simple measurements for This study was financially supported by Foundation Nutrition 2000Plus. Correspondence to: Claude Pichard, MD, PhD, Head, Clinical Nutrition and Diet Therapy, Geneva University Hospital, 1211 Geneva, Switzerland. pichard@cmu.unige.ch Date accepted: January 18, evaluating body composition such as skinfold measurements are easy to perform but not accurate or reproducible. 5 Other methods require subject cooperation (underwater weighing) or sophisticated equipment and skilled technicians (tracer dilution, dual-energy x-ray absorptiometry [DXA], and neutron activation analysis). More recently, bioelectrical impedance analysis (BIA) has been shown to be more accurate for determining leanness or fatness in humans. 6 BIA provides a more reliable measurement of body composition with respect to FFM and FM than does BMI 5 or simpler methods such as skinfolds and height and weight. 7,8 Two previous studies have reported percentiles for FFM, FM, and percentage of FM (%FM). Heitmann 9 reported BIAdetermined 10th and 90th percentiles for FFM, FM, and %FM in a large Danish adult population aged 35 to 65 y. Pichard et al. 10 published BIA-determined percentiles for the same parameters in adults aged 18 to 65 y using different BIA equations for men and women and for obese and non-obese subjects. Although FFM and FM values for adults younger than 64 y are available, no data are available in older subjects, primarily due to a lack of appropriate BIA formulas applicable in elderly subjects and fewer healthy subjects being available. The recent validation of a single BIA equation in subjects 15 to 94 y with BMIs between 17 and 33.8 kg/m 2 now permits the evaluation of FFM and FM in subjects older than 65 y. 11 The aims of this study were to determine reference values for FFM, FM, and %FM by BIA in a large white (Western European) population of healthy subjects in Switzerland, including elderly subjects, observe their differences in age groups of 10 y in those 15 to 98 y, and develop percentile distributions for those parameters. The percentile data presented serve as reference to evaluate devi- Nutrition 17: , /01/$20.00 Elsevier Science Inc., Printed in the United States. All rights reserved. PII S (01)00555-X

82 Nutrition Volume 17, Numbers 7/8, 2001 Reference Values of Fat-Free and Fat Masses in Healthy Subjects 535 TABLE I. ANTHROPOMETRIC AND BIOIMPEDANCE CHARACTERISTICS OF A HEALTHY WHITE POPULATION* Age (y) Men n Height (cm) ( ) Weight (kg) ( ) BMI (kg/m 2 ) ( ) IBW (%) ( ) Resistance ( ) ( ) Reactance ( ) ( ) Phase angle ( ) (degree) Women n Height (cm) ( ) Weight (kg) ( ) BMI (kg/m 2 ) ( ) IBW (%) (70 209) Resistance ( ) ( ) Reactance ( ) ( ) Phase angle (degree) ( ) * Values are presented as mean standard deviation. Ranges are shown in parentheses. BMI, body mass index; IBW, ideal body weight (Metropolitan Life Insurance, 1983) ations from normal values of FFM and FM in healthy adult men and women at a given age. SUBJECTS AND METHODS Subjects Healthy adults (2735 men and 2490 women), aged 15 to 94 y, were recruited non-randomly through advertisements in local newspapers, by offering free BIA on an exhibition stand at trade fairs and fun runs, among public administration staff, and by invitations sent to members of leisure clubs for the elderly to participate in the study. The anthropometric data and number of healthy subjects per age group are shown in Table I. All subjects were ambulatory whites (Western European) who had no known pathologies or physical handicaps. Subjects were questioned on their use of medications and reasons for visiting their physicians within the previous 6 mo to exclude those with acute or chronic diseases, and subjects with water or electrolyte imbalances (e.g., edema and ascites), skin abnormalities (e.g., pachydermia secondary to hypothyrodism), and abnormal body geometry (e.g., amputation and limb atrophy) that might interfere with BIA measurements were excluded. The study protocol complied with the requirements of the Geneva University Hospital Ethics Rules. Anthropometric Measurements and BIA Body height was measured to the nearest 0.5 cm and body weight to the nearest 0.1 kg on a balance-beam scale. Subjects wore indoor clothing without shoes, heavy sweaters, or jackets. One kilogram per subject was deducted for pants and shirt. Percentages of ideal body weights were derived from the 1983 Metropolitan Life Insurance Tables. Values used were the midpoints for medium frame for each height, recalculated in centimeters and kilograms. FFM and FM were assessed by BIA as previously described. 12 Whole-body resistance (R) was measured with four surface electrodes placed on the right wrist and ankle. Briefly, an electrical current of 50 khz and 0.8 ma was produced by a generator (Bio-Z2, Spengler, Paris, France) and applied to the skin using adhesive electrodes (3M Red Dot T, 3M Health Care, Borken, Germany) with the subject lying supine. 13 The skin was cleaned with 70% alcohol. Several BIA instruments were used in this study to permit inclusion of a large number of subjects at one time. The Bio-Z2 generators were cross validated at 50 khz against the RJL-109 and 101 analyzers (RJL Systems, Inc., Clinton Township, MI, USA) and the Xitron 4000B analyzer (Xitron Technologies, Inc., San Diego, CA, USA). The limit of tolerance between instruments for resistance was 5 at 50 khz as determined with a calibration jig and in vivo measurements. The betweeninstrument variability for FFM (n 29) was 0.2 kg (confidence interval of kg), with a coefficient of variation of 1.8%. No difference (P 0.05) was found for resistance at 50 khz between the Xitron, Bio-Z, and RJL 101 device. Earthman et al. 14 reported no significant differences between the Xitron 4000B and RJL 101 devices. Eight observers were used to obtain BIA measurements. All coworkers were trained by the same senior investigator (U.G.K.). In addition, we used a training video to reinforce standarization of our procedure. The interobserver, interday variability (n 51) was 0.02 kg FFM (confidence interval of kg), with a coefficient of variation of 1.3%. FFM was calculated by the following multiple-regression equation with a Xitron 4000B analyzer (Xitron Technologies) that had been previously validated against DXA (Hologic QDR-4500 instrument, Hologic, Inc., Waltham, MA, USA) in 343 healthy subjects aged 18 to 94 y 11 : FFM (0.518 height 2 resistance) (0.231 weight) (0.130 reactance) (4.229 sex [men 1, women 0]) DXA-measured FFM was kg. BIA-predicted FFM was kg, with a bias of kg (r 0.986,

83 536 Kyle et al. Nutrition Volume 17, Numbers 7/8, 2001 TABLE II. PERCENTILES FOR FAT-FREE MASS (kg) BY 50-kHz BIOELECTRICAL IMPENDANCE ANALYSIS FOR HEALTHY WHITE ADULTS Age group (y) n Mean SD Percentile 5th 10th 25th 50th 75th 90th 95th Men All * * Women All * * P 0.01 versus preceding age group, analysis of variance. P versus preceding age group, analysis of variance. standard error of the estimate [SEE] 1.72 kg, technical error [TE] 1.74 kg). Subjects with BMIs above 27 kg/m 2 were analyzed separately to determine whether larger errors occurred in larger subjects with the BIA equation. The mean difference for 67 subjects with BMIs above 27 was kg (P 0.05, paired t-test; r 0.983, SEE 2.0 kg). The BIA equation was further validated against DXA in 205 healthy elderly subjects. 15 We also validated the Bio-Z2 in 250 of the 343 subjects included in the above study (unpublished data) with the BIA equation and found that the FFMs were kg with DXA and kg with the Bio-Z2; the mean difference between DXA and the Bio-Z2 was kg (r 0.99, SEE 1.6 kg; P 0.22, unpaired t-test), which was no different from the bias between DXA and Xitron. Statistics The statistical analysis program StatView, version 4.1 (Abacus Concepts, Berkeley, CA, USA), was used for statistical analysis. The results are expressed as mean standard deviation. Age- and sex-specific percentile distributions were calculated for FFM, FM, and %FM. The data were stratified by intervals of 10 y as reported for BMI and anthropometric data in the NHANES study 16,17 and Canada. 18 The ith percentile (Pi) was the value at or below which there was i% of the sample. For example, the 50th percentile (P50) was the value at or below which were 50% of the observations for a given variable. Given a total of n ordered values for each parameter (x 1, x 2, x 3, x n ), the Pi in any of the calculated distributions was computed as follows: Pi (1 A) (x b ) (A) (x b 1 ). Differences between age groups were analyzed by analysis of variance and Fisher s protected least-significant difference comparison. Statistical significance was set at P 0.05 for all tests. RESULTS Table I shows the anthropometric and bioimpedance characteristics of the men and women with no known pathologies. Age- and sex-specific percentile distributions for FFM, FM, and %FM of white adults are presented in Tables II, III, and IV and Figs. 1, 2, and 3. BMI and Weight Weight and percentage of ideal body weight were lowest in men and women 15 to 25 y old and highest in men and women 65 to 74 y old. Mean BMI was lowest in men and women 15 to 24 y old and highest in men and women 65 to 74 y old and decreased thereafter (Table I). BIA-Derived Body-Composition Parameters Mean FFM (Table II) was greatest in men 35 to 44 y old at kg and decreased thereafter. Of these men, 90% had FFMs between 51.9 and 66.4 kg. Mean FFM was 8.9 kg or 14.8% lower in men older than 85 y than in men 35 to 44 y old. Mean FFM was significantly higher in men 25 to 34 y than in men 15 to 24 y old and significantly lower in men 45 to 54 y old than in men 35 to 44 y old and older than 74 y. The percentiles developed for men 25 to 44 y generally exceeded comparable percentiles calculated for men in younger and older groups. A diagrammatic display of these trends is presented in Fig. 1. For women, the mean FFM was greatest between 45 and 54 y at kg and decreased thereafter. Ninety percent of women had FFMs between 37.0 and 48.0 kg. Mean FFM was 6.2 kg or 14.3% lower in women older than 85 y than in those 45 to 54 y. FFM increased slightly but not significantly in women 15 to 24 y old compared with women 45 to 54 y, decreased thereafter

84 Nutrition Volume 17, Numbers 7/8, 2001 Reference Values of Fat-Free and Fat Masses in Healthy Subjects 537 TABLE III. PERCENTILES FOR FAT MASS (kg) BY 50-kHz BIOELECTRICAL IMPENDANCE ANALYSIS FOR HEALTHY WHITE ADULTS Age group (y) n Mean SD Percentile 5th 10th 25th 50th 75th 90th 95th Men All * * * * Women All * * P versus preceding age group, analysis of variance. P?? versus preceding age group, analysis of variance. TABLE IV. PERCENTILES FOR FAT MASS (%) BY 50-kHz BIOELECTRICAL IMPENDANCE ANALYSIS FOR HEALTHY WHITE ADULTS Age group (y) n Mean SD Percentile 5th 10th 25th 50th 75th 90th 95th Men All * * Women All * * * * P 0.05 versus preceding age group, analysis of variance. P versus preceding age group, analysis of variance.

85 538 Kyle et al. Nutrition Volume 17, Numbers 7/8, 2001 FIG. 1. Percentile changes in fat-free mass (kg) of white adults between the ages of 15 and 98 y. and decreased significantly in women older than 75 y. These age-related trends also were apparent in the percentile distributions. Figure 1 shows these trends in graphic form. FM (Table III) and %FM (Table IV) were lowest in men 15 to 24 y and highest in men older than 85 y. Of these men, 90% had FMs between 8.3 and 22.3 kg and %FMs between 12.6 and The FM increase was significant for all age groups until 74 y, except 45- to 54-y-old versus 35- to 44-y-old men. The mean FM in men older than 85 y was nearly double the FM in men 15 to 25 y old. FM for women (Table III) was lowest in those 15 to 24 y old and highest in those 65 to 74 y old and decreased thereafter. Of these women, 90% had FMs between 10.7 and 25.9 kg. The largest mean FM, observed in women 65 to 74 y, exceeded the smallest mean, noted in women 15 to 24 y, by 51.9%. The mean %FM (Table IV) was lowest in women 15 to 24 y and highest in women 75 to 84 y old. Of these women, 90% had %FMs between 20.8 and A significant increase in %FM was noted only in women 45 to 74 y old. Figure 2 shows trends in FM and Fig. 3 shows trends in %FM FIG. 2. Percentile changes in fat mass (kg) of white adults between the ages of 15 and 98 y.

86 Nutrition Volume 17, Numbers 7/8, 2001 Reference Values of Fat-Free and Fat Masses in Healthy Subjects 539 FIG. 3. Percentile changes in percentage of fat mass (%) of white adults between the ages of 15 and 98 y. for men and women. Percentage of FM increased progressively and percentiles showed parallel distributions in men. In women the percentile-distribution increase did not occur until age 45 y, when considerable increases were noted in all percentiles. FFM and FM in Men and Women As expected, significant differences in FFM (P 0.001) were noted for all age groups between men and women. The absolute differences in FFM between men and women were greatest (17.3 kg) between 25 and 34 y and decreased with age, but the difference remained relatively constant at 36% to 38% throughout that age span. The higher FFM was confirmed by mean FFM indices (FFM/ height 2 ; data not shown) of 19.3 to 19.4 kg/cm 2 in men 35 to 74 y old ( kg/cm 2 in men younger than 35 y and kg/cm 2 in men older than 74 y) compared to 16.2 to 16.5 kg/cm 2 in women 45 to 74 y old ( kg/cm 2 in women younger than 45 y and kg/cm 2 in women older than 74 y). Sex differences also were noted in FM and %FM. FM was significantly higher in women than in men, which was confirmed by the higher mean FM indices. The FM index increased with advancing age, from kg/cm 2 to kg/cm 2 in women and from kg/cm 2 to kg/cm 2 in men. Mean %FM was significantly greater (P ) in women than in men. Furthermore, larger FM and %FM in older adults showed that any weight gain was explained by FM gains in both sexes. DISCUSSION Few studies have reported FFM and %FM in large population samples. 5,19,20 In general, FM is derived by indirect methods. The development and validation of BIA now permit the determination of FFM and FM more accurately by an easy, portable, and inexpensive method. 5,8 Two previous studies have reported percentiles for FFM, FM, and %FM by BIA in subjects 15 to 65 y old 9,10 but did not evaluate men and women older than 65 y. The aims of this study were to determine reference values for FFM, FM, and %FM in a white population of healthy adult subjects, observe their differences with age, and develop percentile distributions for those parameters between the ages of 15 and 98 y. Application of BIA Formulas in Subjects With Different Ages and BMIs The validity of BIA-determined FFM and FM directly depends on the equation used to translate the BIA-determined resistance and reactance values into FFM, FM, or total body water. 5,21,22 FFM in the present study was estimated with a multiple regression equation that was previously validated against DXA in 343 healthy subjects 18 to 94 y old. 11 Validation of the Geneva BIA equation was deemed necessary because published equations were inadequate for overweight, obese, and elderly subjects. Non-significant differences were found in subjects older than 65 y and obese subjects. 11 Inclusion of reactance in the BIA equation allowed the distinction of lower FFMs in older subjects and improved correlation SEE and TE compared with other BIA equations. Comparison of BMI With Previous Studies The BMIs in the present study can be compared with those in other epidemiologic studies. Median BMIs of 24.6 to 25.1 kg/m 2 in 35- to 74-y-old men and 21.8 to 23.6 kg/m 2 in 35- to 74-y-old women reported in a randomly sampled Swiss population 23 were similar to those in the current study. In other studies, mean BMIs were kg/m 2 in men and 25.3 kg/m 2 in women aged 15 to 99 y in Finland, 24 and median BMIs were 24.5 to 25.3 kg/m 2 in men and 22.3 to 24.0 kg/m 2 in women 35 to 65 y old in France; subjects in both studies were selected randomly. 25 By definition, % of men and 14.5% of women were overweight (BMI 25.0 kg/m 2 ) and 2.9% of men and 3.8% of women were obese (BMI 30 kg/m 2 ) in the present study. In comparison, 34% to 35% of men and 24% to 25% of women in Britain, Netherlands, and Australia are overweight, as are 38% to 40% of men and 21% to 22% of women in the United States and Canada. 24 Although some underrepresentation of subjects with BMIs above 25 is possible, due in part to excluding subjects with secondary pathologies, the results suggest that prevalence of overweight and obesity is lower in Switzerland than in other European countries and considerably lower than in the United States and Canada.

87 540 Kyle et al. Nutrition Volume 17, Numbers 7/8, 2001 Evaluation of Body-Composition Variations During Aging FAT-FREE MASS. In our study, the FFM peaked in men 35 to 44 y old and women 45 to 54 y old and declined thereafter, compared with declines beginning at age 60 y in men and 45 y in women in the study by Bartlett at al. 27 Heitmann 9 found that FFMs were highest in 35-y-old men and women and decreased thereafter. Forbes 28 suggested that a weight gain of 2.3 kg/decade is necessary to avoid losing FFM during aging. The low body weight increases reported in our subjects (approximately 5 kgin40y) would have been inadequate to maintain FFM during aging and would explain the earlier decline of FFM in our subjects. Heitmann 9 also reported small differences in weight between age groups ( 1.4 kg in men and 2.1 kg in women between 35 and 55 y), which also would have been inadequate to maintain FFM during aging. In contrast, the Fels Longitudinal Study 29 reported stable FFMs and larger weight gains in men and women 20 to 59 y than in our study, which could explain the discrepancies between the FFM values. We found lower FFMs in men and women older than 60 y and an accelerated loss in men and women older than 75 y. Most studies have reported lower FFMs after age 60 y. 29,30 In a crosssectional study of men and women older than 60 y, Baumgartner also found the weight, BMI, and FFM decreased with age. Forbes 28 did not find an accelerated loss in FFM in a small longitudinal study but did find an average loss of 1.5 kg of FFM/decade in subjects who maintained their weight. Thus, the accelerated decrease in our subjects who were older than 75 y is probably related to lower weight noted after age 75 y compared with weight maintenance or increase until age 74 y. In our study, linear regressions suggested declines of 1.5 and 1.2 kg of FFM/ decade in men and women, respectively, 55 to 98 y old and 1.9 and 1.8 kg of FFM/decade in men and women, respectively, older than 75 y, with weights being progressively lower in older subjects. Longitudinal studies are necessary to determine changes in FFM with aging in view of changes in weight. Thus, the differences in FFM between studies can be explained by differences in simultaneous weight changes, and FFM might increase, remain stable, or decrease between the ages of 20 and 60 y depending on weight gain or loss. FFM does appear to decrease after age 60 y, probably because weight gains are no longer large enough to offset the inevitable loss of FFM with aging. FAT MASS AND PERCENTAGE OF FAT MASS. Our study showed that body weight and BMI increased until age 74 y and that the increase was predominantly due to higher FM. FM continued to increase in men after age 74 y, and %FM increased throughout the lifespan in men and women. 24,31 This study agrees with previous reports and shows higher FM and %FM in men in all age ranges, which is parallel to higher weights. Because the data are cross-sectional rather than longitudinal, trends in body composition parameters observed with advancing age might have represented differences between successive generations in addition to physiologic alterations with aging. Linear regressions in subjects between 15 and 98 y in our data showed %FM increases of 1.5/decade in men and 1.7/decade in women compared with 2.3/decade in the study by Deurenberg. 39 Confirmation of an effect of age on FM, in addition to an effect of weight gain with age, by longitudinal studies is necessary. Higher %FMs have been reported in a number of studies that included subjects with higher BMIs. 5,20,29,32 Roubenoff et al. 5 found a mean increase in %FM with age in both sexes that peaked in the fifth and sixth decades in American women and men, which is inconsistent with our results of increases in %FM until our oldest age groups in men and women. Biasoli et al. 33 found that %FM increased until age 90 y. Guo et al. 4 reported higher FM and %FM by underwater weighing in men and women aged 40 to 66 y who had BMIs higher than those reported in our study and found significant decreases in FFM and increases in total FM, %FM, body weight, and BMI with age in their longitudinal study (mean follow-up 9 y). Thus, changes in body composition with age need to be confirmed with longitudinal studies. Comparisons between studies where height, weight, and age are significantly different are difficult to make. We could not confirm a highly significant curvilinear relation between age and FM, which indicated a peak amount of FM in late middle age and lower amounts at younger and older ages, as noted by Mott et al. 34 Our data showed a linear relation between age, FM, and %FM. The continuation of FM gain until approximately 75 y might have been due to our inclusion of only healthy subjects. Our selection process might have excluded those individuals who had disease-related declines in weight, FFM, and FM, which would have contributed to the age-related declines. Comparisons of Percentiles With Previous Studies The P50s for FM were 2.9 kg higher in men and 2.0 kg higher in women than in our previous study. 10 The discrepancy stems from an underestimation of FM by the BIA equations used in the previous study. Recent validation of a single BIA equation against newer versions of DXA hardware and software allows more accurate predictions of FM and FFM and can be used to follow people longitudinally during aging and when significant BMI changes occur. Although the median BMIs reported by Heitman 9 in a Danish population were not remarkably different from those in our study, median FFMs were higher (61.8 kg in 35-y-old and 56.4 kg in 65-y-old men and 45.6 kg in 35-y-old and 41.2 kg in 65-y-old women) in the taller and heavier subjects. Differences were greatest in the P90: weight was 9.0 kg, FFM was 2.6 kg, and FM was 6.2 kg higher in Danish than in Swiss 35-y-old men, and weight was 9.2 kg, FFM was 2.4 kg, and FM was 5.5 kg higher in Danish than in Swiss 35-y-old women. Although differences between the two populations were lower in those aged 65 y, e.g., weight was 4 kg higher, FFM was identical, and FM was 4 kg higher in Danish men and weight was 3.4 kg, FFM was 0.9 kg, and FM was 3.7 kg higher in Danish than in Swiss women, the differences remained considerable. These differences in results suggest that, to adequately evaluate individuals, reference data describing local levels and patterns are necessary because ethnic differences in height, weight, and BMI and differences in lifestyle, environment, and genetic makeup exist between countries and would influence interpretation of FFM, FM, and %FM. 35 Ideal Fat Mass A recent round-table discussion 36 suggested that the best FM percentages in terms of lowest morbidity and mortality averaged 12% to 20% in men and 20% to 30% in women. Forty-five percent of all men and 38% of all women in our study were above those recommendations (Table V). High FMs, defined as the P85, in this study were 25.6% in men and 35.7% in women, and excess FMs, defined as the P95, were 29.2% in men and 40.5% in women. Further research is necessary to determine to what extent health outcomes are affected in Swiss and other populations by the much-higher-than-recommended quantities of FM in individuals. 4,9,37

88 Nutrition Volume 17, Numbers 7/8, 2001 Reference Values of Fat-Free and Fat Masses in Healthy Subjects 541 Study Limitations The subjects in this study were not randomly selected. However, the BMIs in our study were similar to the median BMIs of 24.6 to 25.1 kg/m 2 in 35- to 74-y-old men and 21.8 to 23.6 kg/m 2 in 35- to 74-y-old women in a randomly sampled population in Geneva. 23 Subjects with extreme body FM are frequently underrepresented in non-randomly selected populations. Twenty-four percent of subjects had BMIs above 25 kg/m 2 compared with 29.2% reported in the 4th Nutrition Report in Switzerland (n 510). 38 This difference is probably due to a higher percentage of younger subjects being included in our study than in the Nutrition Report. The subjects in this study were volunteers in good health and might not have been representative of the general population. Regular physical activity (walking) and the absence of mobility problems appear to have aided in maintaining physical functioning and might have limited the loss of FFM in men and women older than 55 y. Results in subjects older than 85 y must be interpreted with caution because of the small number of subjects. Percentiles at the 10th and 90th levels in those older than 65 y might be less certain because samples included fewer than the approximately 250 subjects needed 2 to calculate precise percentiles from the 5th to 95th levels. This study used single-frequency BIA methodology and therefore did not measure total body water, intra- and extracellular water, and their influence on FFM composition with aging. Further research is necessary to explore the extent to which changes in extracellular fluid occur and influence changes in FFM and body cell mass with aging. CONCLUSION Mean FFM peaked in men 35 to 44 y old and women 45 to 54 y old and declined thereafter. Mean FM and %FM increased progressively in men and women throughout the ages studied. The results suggested that the higher weight noted in older subjects is due to higher FM. The percentile data serve as references to evaluate deviations from normal values of FFM and FM in healthy adult men and women at a given age. ACKNOWLEDGMENTS The authors are indebted to the dietitians at the Geneva University Hospital for data collection. REFERENCES 1. Baumgartner RN, Koehler KM, Gallagher D, et al. Epidemiology of sarcopenia among the elderly in New Mexico. Am J Epidemiol 1998;147: Chumlea WC, Vellas B, Guo SS. Malnutrition or healthy senescence. Proc Nutr Soc 1998;57: Heymsfield SB, Gallagher D, Poehlman ET, et al. Menopausal changes in body composition and energy expenditure. Exp Gerontol 1994;29: Guo SS, Zeller C, Chumlea WC, Siervogel RM. Aging, body composition, and lifestyle: the Fels Longitudinal Study. Am J Clin Nutr 1999;70: Roubenoff R, Dallal GE, Wilson PWF. Predicting body fatness: the body mass index vs estimation by bioelectrical impedance. Am J Public Health 1995;85: Kyle UG, Pichard C. Dynamic assessment of fat-free mass during catabolism and recovery. Curr Opin Clin Nutr Metab Care 2000;3: Heitmann B. Impedance: a valid method in assessment of body composition. Eur J Clin Nutr 1994;48: Bioelectrical impedance analysis in body composition measurement: National Institutes of Health Technology Assessment Conference Statement. Am J Clin Nutr 1996;64:524S 9. Heitmann BL. Body fat in the adult Danish population aged years: an epidemiological study. Int J Obes 1991;15: Pichard C, Kyle UG, Bracco D, et al. Reference values of fat-free and fat masses by bioelectrical impedance analysis in 3393 healthy subjects. Nutrition 2000;16: Kyle UG, Genton L, Karsegard L, Slosman DO, Pichard C. Single prediction equation for bioelectrical impedance analysis in adults aged yrs. Nutrition 2001;17: Lukaski HC. Validation of tetrapolar bioelectrical impedance measurements to assess human body composition. J Appl Physiol 1986;60: Houtkooper LB, Lohman TG, Going SB, Howell WH. Why bioelectrical impedance analysis should be used for estimating adiposity. Am J Clin Nutr 1996;64: 436S 14. Earthman CP, Matthie JR, Reid PM, et al. A comparison of bioimpedance methods for detection of body cell mass change in HIV infection. J Appl Physiol 2000;88: Genton L, Karsegard VL, Kyle UG, et al. Comparison of four bioelectrical impedance analysis formulas in healthy elderly subjects. Gerontology 2001(in press) 16. Micozzi MS, Albanes D, Jones DY, Chumlea WC. Correlations of body mass indices with weight, stature, and body composition in men and women in NHANES I and II. Am J Clin Nutr 1986;44: Frisancho R. New standards of weight and body composition by frame size and height for assessment of nutritional status of adults and the elderly. Am J Clin Nutr 1984;40: Macdonald SM, Reeder BA, Chen Y, Despres JP. Obesity in Canada. a descriptive analysis. Canadian Heart Health surveys Research Group. Can Med Assoc J 1997;157:S3-S9 19. Micozzi MS, Harris TM. Age variations in the relation of body mass indices to estimates of body fat and muscle mass. Am J Phys Anthropol 1990;81: Bartlett HL, Puhl SM, Hodgson JL, Burskirk ER. Fat-free mass in relation to stature: ratios of fat-free mass to height in children, adults, and elderly subjects. Am J Clin Nutr 1991;53: Pichard C, Kyle UG. Body composition measurements during wasting diseases. Curr Opin Clin Nutr Metab Care 1998;1: Pichard C, Kyle U, Gremion G, Gerbase M, Slosman D. Body composition by X-ray absorptiometry and bioelectrical impedance in elite female runners. Med Sci Sports Exerc 1997;29: Morabia A, Bernstein M, Héritier S, Ylli A. Community-based surveillance of cardiovascular risk factors in Geneva: methods, resulting distributions, and comparisons with other populations. Prevent Med 1997;26: Rissanen A, Heliövaara M, Aromaa A. Overweight and anthropometric changes in adulthood: a prospective study of Finns. Int J Obes 1988;12: Rolland-Cachera MF, Cole TJ, Sempé M, et al. Body mass index variations: centiles from birth to 87 years. Eur J Clin Nutr 1991;45: World Health Organisation. Diet, nutrition and the prevention of chronic disease. Geneva: World Health Organisation, Barlett HL, Puhl SM, Hodgson JL, Buskirk ER. Fat-free mass in relation to stature: ratios of fat-free mass to height in children, adults and elderly subjects. Am J Clin Nutr 1991;53: Forbes GB. Longitudinal changes in adult fat-free mass influence of body weight. Am J Clin Nutr 1999;70: Chumlea WC, Guo SS, Zeller CM, Reo NV, Siervogel RM. Total body water data for white adults 18 to 64 years of age: the Fels Longitudinal Study. Kidney Int 1999;56: Mazariegos M, Wang ZM, Gallagher D, et al. Differences between young and old females in the five levels of body composition and their relevance to the two-compartment chemical model. J Gerontol 1994;49:M Bishop C, Phyllis E, Ritchey S. Norms for nutritionnal assessment of American adults by upper arm anthropometry. Am J Clin Nutr 1981;34: Kehayias JJ, Fiatarone MA, Zhuang H, Roubenoff R. Total body potassium and body fat: relevance to aging. Am J Clin Nutr 1997;66: Biasioli S, Foroni R, Petrosino L, et al. Effect of aging on the body composition of dialyzed subjects. Comparison with normal subjects. ASAIO J 1993;39:M Mott JW, Wang J, Thornton JC, et al. Relation between body fat and age in 4 ethnic groups. Am J Clin Nutr 1999: de Onis M, Habicht JP. Anthropometric reference data for international use: recommendations from a World Health Organization Expert Committee. Am J Clin Nutr 1996;64: Abernathy RP, Black DR. Healthy body weights: an alternative perspective. Am J Clin Nutr 1996;63:448S 37. Baumgartner RN. Body composition in healthy aging. Ann NY Acad Sci 2000; 904: Paccaud F, Wietlisbach V, Rickenbach M. Evolution des maladies cardiovasculaires et des caractéristiques de l alimentation: résultats de l étude MONICA. In: Office Fédéral de la Santé Publique, ed. Quatrième rapport sur la nutrition en Suisse. Bern: Office Fédéral de la Santé Publique, 1998: Deurenberg P, Westrale JA, Seidell JC. Body mass index as a measure of body fatness: age- and sex-specific prediction formula. Brit J Nutr 1991;65:105

89 APPENDIX 6

90 ARTICLE IN PRESS Clinical Nutrition (2004) 23, ORIGINAL ARTICLE Aging, physical activity and height-normalized body composition parameters Ursula G. Kyle a, Laurence Genton a,g!erald Gremion b, Daniel O. Slosman c, Claude Pichard a, * a Division of Clinical Nutrition, Geneva University Hospital, 1211 Geneva, Switzerland b Division of Orthopedics, Lausanne University Hospital, Lausanne, Switzerland c Division of Nuclear Medicine, Geneva University Hospital, Geneva, Switzerland Received 1 October 2002; accepted 22 May 2003 KEYWORDS Fat-free mass; Body fat; Body composition; Gender; Exercise; Aging Summary Background & Aim: Regular physical activity prevents or limits weight gain, and gain in body mass index (BMI) and decreases mortality. The aims of the study in healthy adults were to determine the differences in fat-free mass index (FFMI) (kg/ m 2 ) and body fat mass index (BFMI) between age groups and determine the association between physical activity and FFMI and BFMI. Methods: Caucasian men (n ¼ 3549) and women (n ¼ 3184) between 18 and 98 years, were classified as either sedentary or physically active (at least 3 h per week at moderate or high-intensity level activity). FFMI and BFMI were measured by 50 khz bioelectrical impedance analysis. Results: BFMI was significantly higher (Po0.05) in sedentary than physically active subjects and the differences became progressively greater with age. The physically active subjects were significantly less likely to have a low or high FFMI (logistic regression, Po0.001), and a high or very high BFMI (Po0.001), and more likely to have low BFMI (Po0.001) compared to sedentary adults. In contrast with fat-free mass, which was lower in older subjects, the height-normalized FFMI was stable with age until 74 years and lower thereafter. Significantly higher BFMIs were noted in older subjects. Conclusion: Physically active subjects are less likely to have low or high FFMI, and high or very high BFMI, and more likely to have low BFMI. In contrast to common claim that fat-free mass decreases with age, we found that FFMI was stable until 74 years. The use of FFMI and BFMI permits comparison of subjects with different heights and age. & 2003 Elsevier Ltd. All rights reserved. Introduction *Corresponding author. Tel.: þ ; fax: þ address: claude.pichard@medecine.unige.ch (C. Pichard). Overweight and obesity are clearly associated with increased mortality 1,2 and chronic diseases. 3 Low physical activity is associated with increased allcause mortality rates, 4,5 and regular physical activity prevents or limits weight gain, and gain in S /$ - see front matter & 2003 Elsevier Ltd. All rights reserved. doi: /s (03)00092-x

91 ARTICLE IN PRESS 80 U.G. Kyle et al. body mass index (BMI). 6,7 However, BMI is an imprecise measurement of fatness. 8 Recent findings point out the importance of measuring body composition in clinical programs aimed at influencing disease prevention 9 and suggest that body composition can be used to assess the body fat (BF) and fat-free mass. Segal et al. 10 demonstrated that high BF was associated with hypertension and diabetes. A high percentage of BF was also significantly associated with an increase in total mortality, compared with a low percentage of BF. 11 Heitmann et al. 12 confirmed that total mortality was a linear increasing function of low fat-free mass and high BF. Fat-free mass and BF vary with body height, age and physical activity. It therefore seems inappropriate to give, for any individual, a cut-off point in absolute value (kg) below or above which fat-free mass or BF are judged as low or high, respectively. For example, a short individual would be penalized since his absolute fat-free mass is expected to be lower than that of a tall individual. Therefore, in analogy to the use of the BMI (weight/height 2 ) for grading relative adiposity, the relative values of fat-free mass and BF, namely fat-free mass index (FFMI) and body fat mass index (BFMI) (kg/m 2 ), permit comparison of subjects with different height. Our recent population studies show that patients at hospital admission had significantly lower FFMI and higher BFMI than healthy subjects with similar BMI. 13 In addition, low FFMI and high BFMI were associated with increased length of stay in these patients. 14 This strongly suggests that FFMI and BFMI are better at discriminating body compartments than BMI. On this basis, we extended our previous work, which determined the association between sedentarism and physical activity in 3853 adults aged less than 65 years 7 by evaluating the body heightnormalized parameters in 6733 subjects and including elderly over 65 years. The aims of this study were to describe the differences in FFMI and BFMI between age groups and determine the association between physical activity and FFMI and BFMI in healthy Caucasian subjects aged 18 and 98 years. Subjects and methods Subjects Healthy Caucasian adults (n ¼ 3549 men and 3184 women), aged years, were recruited by offering free BIA measurements at trade fairs, leisure clubs and fun runs, and among public administration staff and their relatives. The subjects represented a wide range of physical activity levels. Subjects were questioned on use of medications, and reasons for visit to physician in the last 6 months to eliminate subjects with acute or chronic diseases. Subjects with known acute pathologies or physical handicap were excluded. Smoking was not an exclusion criterion. The study protocol complied with the requirements of the Geneva University Hospital Ethics Rules. Written informed consent was obtained from all subjects. Physical activity Subjects completed a questionnaire to specify the hours and minutes of physical activity per week performed on a regular basis throughout the year. The seasonal variations and the type of activity were determined by the questionnaire. The purpose of the questionnaire was to discriminate active from sedentary subjects. The subjects who performed 43 h of physical activity per week for longer than 2 months were classified as physically active. Only physical activity at moderate or high intensity (4.0 METS or more, as defined by the activity intensity codes by the Minnesota Leisure Time Activities Questionnaire 15 ), was counted. All others were classified as sedentary. Resistance type activity accounted for o10% of overall activity, thus could not be analyzed separately. Home- and work-related physical activity was not considered in this study. Our decision not to include home activity was based on the assumption that activity would not be different between sedentary and physically active subjects. We therefore limited physical activity to activities that were intended for recreational and training purposes. Misclassifications of some subjects were possible, namely some physically active subjects might not have been recognized as physically active because of missing data and would have been included in the sedentary group. However, this could have caused the mean BF mass to be lower in the sedentary group, and thus might have reduced the BFMI differences between Physically Active and Sedentary group. The misclassification would not have affected the conclusions (see limitations of the study). Anthropometric measurements and bioelectrical impedance analysis Body height was measured to the nearest 0.5 cm and body weight to the nearest 0.1 kg on a balance beam scale. Subjects were in indoor clothing

92 ARTICLE IN PRESS Aging, physical activity and body composition 81 without shoes and heavy sweaters or jackets. One kilogram was deducted for pants and shirt. Whole-body resistance (R) and reactance were measured with four surface electrodes placed on the right wrist and ankle, as previously described and taking standard precautions for BIA measurements. 16 Subjects were recumbent for 45 min and did not perform physical activity prior to the BIA measurement. Briefly, an electrical current of 50 khz and 0.8 ma was generated (Bio-Z2 s, Spengler, Paris, France) and applied to the skin using adhesive electrodes (3 M Red Dot T, 3 M Health Care, Borken, Germany) with the subject lying supine. 17 The skin was cleaned with 70% alcohol. Fat-free mass was calculated as follows by previously validated (n ¼ 343) BIA equation 18 : Fatfree mass ¼ þ (0.518 height 2 /resistance) þ (0.231 weight) þ (0.130 reactance) þ (4.229 sex (men ¼ 1, women ¼ 0)). All data collectors were trained to standardize the data collection. All measurements of participants in fun runs were made prior to their runs, to avoid problems of changes in hydration, skin temperature, electrolyte concentration, and glycogen stores. FFMI was calculated as FFMI (kg/m 2 ) ¼ fat-free mass (kg) divided by height 2 (m 2 ) and BFMI ¼ BF (kg) divided by height 2 (m 2 ). Ranges for low, normal and high (for BMI kg/m 2 ) FFMI and low, normal, high and very high BFMI categories (Appendix A) were derived from our healthy subjects (n ¼ 5635) 19 from polynomial regression equations for each of the BMI cutoffs (18.5, 25, 30 kg/m 2 ), which correspond to World Health Organization categories for low weight, normal weight, overweight and obese. Statistical analysis The results are expressed as mean7standard deviation (x7sd). The data were stratified by steps of 10 years as previously reported. Unpaired t-tests were used (Statview s 5.0, Abacus Concepts, Berkeley, CA) to compare BMI, fat-free mass, FFMI, BF, BFMI and % BF between sedentary and physically active subjects. ANOVA was used to determine significant changes between different age groups for each of the body composition parameters, with Fisher Protected Least Significant Difference comparison used to determine individual significant differences. Men and women and sedentary and physically active subjects were evaluated separately. Multiple regressions were used to determine the effects of weight, height, gender, activity and age on FFMI and BFMI. Gender was coded as men ¼ 1, women ¼ 0; activity level as physically active ¼ 1, sedentary ¼ 0. Two-way analysis of variance was used to evaluate the overall effects of age group, physical activity and gender on FFMI and BFMI. Statistical significance was set at Po0.05 for all tests. Results The effects of physical activity could not be evaluated in men over 85 years and women over 75 years, because subjects in these age groups no longer met physical activity criteria. Effects of physical activity on fat-free mass index and body fat mass index Weight and BMI (Table 1) were significantly higher (Po0.05) in sedentary than in physically men and women between 25 and 74 years. Similar observations were made for BF. The differences in BF between physically active and sedentary men and women became progressively larger between 35 and 64 years. FFMI (Table 2) was slightly higher in sedentary than physically active men and women. BFMI (Table 2) was significantly higher (Po 0.05) in sedentary than physically active men and women, and the differences became progressively greater with age. The physically active subjects (men and women combined) were significantly less likely to have a low or high FFMI (Po0.001) and a high or very high BFMI (Po0.001), and more likely to have low BFMI compared to sedentary adults, as indicated by significant odd ratios (OR) (Po0.001) (Table 3, Figs. 1 and 2). The ORs did not differ between men and women and were therefore combined. Thus physical activity resulted in lower prevalence of low FFMI, but did not result in higher prevalence of high FFMI. It also resulted in lower prevalence and risk of having high and very high BF in physically active compared to sedentary men and women (Table 3). Effects of age on body composition The BMI was 15% and 14%, 23% and 12% higher in years than in years sedentary and physically active men and women, respectively. In contrast, BFMI was 71% and 66% and 68 and 29%, respectively, higher in the same subjects groups. The FFMI was 4 and 4% higher in sedentary and physically active men, 6 and 7% higher women, in y than in y sedentary and physically

93 ARTICLE IN PRESS 82 U.G. Kyle et al. Table 1 Anthropometric characteristics of 6733 healthy Caucasian sedentary and physically active adults. Age Height Weight BMI (years) Sedentary Physically active Sedentary Physically active Sedentary Physically active n (kg) n (kg) (kg) (kg) (kg/m 2 ) (kg/m 2 ) Men w * y y,z y y,z *,w * z y z y y z y *,z * y z z *,w w * * *,w * y NA NA Women * z * z y y z y *,z * y z y *,z y * z y z * * * z y w * NA * NA y * NA NA Mean7SD; ANOVA comparison to preceeding age group: *Po0.05, y Po Unpaired t-test comparison of sedentary vs. physically active: w Po0.05, z Po0.001, significance level Po0.05. NA ¼ not available. active men and women, respectively. Thus in contrast with fat-free mass, which was lower in older men and women, the FFMI remained stable with age until 74 years, then declined. Significantly, higher BFMIs were noted in older who had higher BMI than younger subjects. Combined effects of physical activity, age and gender FFMI was 14 18% lower in women than men, depending on age. Positive trend tests for FFMI (Table 4), were noted for weight, gender and activity (Po0.001) and negative trends for height and age (Po0.001). Gender indicated that FFMI would be 2.8 kg/m 2 higher and BFMI 2.8 kg/m 2 lower in men than in women. Activity increased FFMI by 0.32 and decreased BFMI by 0.34 kg/m 2 (men and women combined). Age decreased FFMI and increased BFMI by 0.02 kg/m 2 /yr. Furthermore, the effect of age on FFMI was and 0.02 kg/ m 2 /yr in physically active and sedentary men and women, respectively. A two-way ANOVA showed an effect of physical activity, age and gender on FFMI and BFMI (Po0.001). These results suggest that the physical activity had an effect on BF, but did not affect fatfree mass. Discussion Our study shows that physically active subjects are less likely to have low or high FFMI, high or very high BFMI, and more likely to have low BFMI than sedentary subjects. The use of FFMI and BFMI permits comparison of subjects with different height and age. In contrast to common claim that fat-free mass decreases with age, we found that FFMI was stable until 74 years, but the BFMI was higher in older compared to younger subjects. Effects of physical activity on body composition Physically active men and women had lower weight, BMI and % BF. The lower weights and BMIs in physically active subjects resulted in a lower prevalence of high FFMI compared to sedentary subjects (Fig. 1). Our study also shows that

94 ARTICLE IN PRESS Aging, physical activity and body composition 83 Table 2 Body composition parameters of 6733 healthy Caucasian sedentary and physically active adults. Age Fat-free mass Body fat Fat-free mass index Body fat mass index (years) Sedentary Physically Sedentary Physically Sedentary Physically Sedentary Physically active active active active (kg) (kg) (%) (%) (kg/m 2 ) (kg/m 2 ) (kg/m 2 ) (kg/m 2 ) Men w * w y y,z * y y y,z z y z y z *,z y *,z * w y *,z þ * * z * z *,w y,w y,w y z y w y NA * NA NA * NA Women w w z * z w * z y y,w * z * z y z * w y z * y *,z w y *,z * y z y z y NA NA y NA NA y NA NA NA NA Mean7SD; ANOVA comparison to preceeding age group: *Po0.05, y Po Unpaired t-test comparison of sedentary vs. physically active: w Po0.05, z Po0.001, significance level Po0.05. NA ¼ not available. Table 3 Odds ratio (95% CI), adjusted for age, of low, normal and high fat-free mass index (FFMI) and low, normal, high and very high body fat mass index (BFMI) for sedentary versus physically active men and women (n ¼ 6733). %(n) % (n) OR (95% CI) FFMI Low Normal Sedentary 8.1 (310) 58.1 (2222) 1 Physically active Low w 4.8 (140) 66.8 (1941) 0.7 ( ) n High Normal Sedentary 33.8 (1294) 58.1 (2222) 1 Physically active High w 28.4 (826) 66.8 (1941) 0.8 ( ) n BFMI Low Normal Sedentary 4.2 (162) 60.7 (2324) 1 Physically active Low w 6.8 (198) 73.3 (2131) 2.0 ( ) n High Normal Sedentary 27.7 (1060) 60.7 (2324) 1 Physically active High w 18.6 (542) 73.3 (2131) 0.4 ( ) n Very high Normal Sedentary 7.3 (280) 60.7 (2324) 1 Physically active Very high w 1.2 (36) 73.3 (2131) 0.1 ( ) n OR ¼ odds ratio, CI ¼ confidence interval, logistic regressions, adjusted for age and gender. n Po w See methods for classifications.

95 ARTICLE IN PRESS 84 U.G. Kyle et al. Figure 1 Prevalence (%) of fat-free mass index in sedentary and physically active healthy Caucasian adults. Prevalence (%) of low, normal and high fat-free mass index by age group in sedentary (left) and physically active women (top) and men (bottom). Physically active men and women were more likely to have normal fat-free mass index and less likely to have high fat-free mass index than sedentary men and women. Figure 2 Prevalence (%) of body fat mass index in sedentary and physically active healthy Caucasian adults. Prevalence (%) of low, normal, high and very high body fat mass index by age group in sedentary (left) and physically active women (top) and men (bottom). Physically active men and women over 35 years of age were less likely to have high and very high body fat mass index than sedentary men and women. Physically active women were also more likely to have low fat mass index than sedentary women. physically active subjects were less likely to have low FFMI. Physical activity was predominantly endurance rather than resistance type and would not necessarily have increased fat-free mass. 20 Physical activity also resulted in a lower prevalence of high and very high BFMI in physically active compared to sedentary subjects (Table 3, Fig. 2). It further showed a higher prevalence and significant OR for having low BFMI. Inactivity has been shown to contribute to overweight and obesity. Subjects with low physical activity at follow-up were 3.8 times more likely to

96 ARTICLE IN PRESS Aging, physical activity and body composition 85 Table 4 Multiple-regression results for fat-free mass index (FFMI) and body fat mass index (BFMI) on independent variables in 6733 adults. Regression CI w Po Regression CI w Po coefficient n coefficient n FFMI (kg/m 2 ) BFMI (kg/m 2 ) Intercept ; ; Weight (kg) ; ; Height (cm) ; ; Gender z ; ; Activity y ; ; Age (years) ; ; r ¼ 0.92 r 2 ¼ r ¼ 0.93 r 2 ¼ n 7Standard error. w CI ¼ 95% confidence interval. z Gender: men ¼ 1, women ¼ 0. y Activity: sedentary ¼ 0, physically active ¼ 1, significance level Po0.05. have gained 413 kg during the preceding 10 years. 21 Walking more than 4 h/week resulted in a modest decrease in BMI and small protection against gain at the waist. 22 We found lower BMI and BF in physically active subjects who exercised for at least 3 h/week. Highest body mass gain was noted in subjects who decreased their activity and was lowest in subjects who increased their activity. 23 Differences in fatness between physically active and inactive women were already present at the age of 25 years and persisted throughout the adult life, 24 an observation confirmed by our study. Exercise-induced weight loss and age-related weight gain appear to have opposing additive effects. 25 Middle age runners are leaner than sedentary men, not because the processes that promote age-related weight gain are abated, but rather because exercise-induced weight loss offsets weight gain during middle age. A strong inverse association of running distance with BMI and waist circumference occurred in all age groups, 25 supported in our study by lower BMI and BF in physically active than sedentary subjects which confirms the association of physical activity and lower BF. Furthermore the age-related decrease in FFMI is lower in physically active than sedentary subjects ( and 0.02 kg/m 2 /year, respectively). Effects of gender and age on body composition FFMI was 14 18% lower in women than men. Multiple regression analysis (Table 4) showed a significant positive effect ( þ 2.8 kg) of gender (men versus women) on FFMI and negative effect ( 2.8 kg) on BFMI. The higher FFMIs in sedentary subjects aged years, who had higher weights and BMIs than subjects o34 years, support the hypothesis that more lean tissue (fat-free mass) is necessary to support higher body weight, 26 For BMIs above 27.4 kg/m 2, higher weights may exert trophic effects in addition to activity effect. Our study supports finding by others 19,21,23 of lower fat-free mass and higher BF in older subjects (Table 2). However, the height-normalized fat-free mass, expressed as FFMI, was lower only in 475 years than in younger subjects. The FFMI was not lower in older subjects for 2 reasons: (1) normalization of fat-free mass for height eliminates a large part of the differences due to height between young and old; (2) because of weight gain, older subjects have a higher BMI and it is know that any weight gain is accompanied by a gain in both BF and fat-free mass. Thus the older subjects are able to maintain their FFMI, because the weight and fatfree mass gain during middle age offset a part of any age-related decrease in FFMI. FFMI was lower only in over 75 years compared to younger subjects. The physically active subjects do not have lower FFMI than younger subjects, in spite of lower weight and BMI gain during middle age, because the age-related decrease in FFMI is lower in physically active than sedentary subjects ( and 0.02 kg/m 2 /yr, respectively). We also found progressively higher BF in all age groups, but the differences were lower in physically active than sedentary men and women.

97 ARTICLE IN PRESS 86 U.G. Kyle et al. Multiple regressions show a positive effect of weight, being male, activity and a negative effect of height and age on FFMI. The significant, though small, effect of age on FFMI confirms that fat-free mass decreases with age. The FFMI decrease with age of 0.02 kg/m 2 /years would result in a 0.8 kg/m 2 lower FFMI in 75 years than 35 years old men or women. The greater positive effect of weight on FFMI than the negative effect of age explains the stable FFMI in men and women until 74 years. FFMI would be maintained as long as the weight increase was sufficient to counteract the small decrease of FFMI with age. Our results suggest that a 1.4 and 0.9 kg weight gain/decade of age in sedentary and physically active adults would result in maintenance of fat-free mass. This is lower than a weight gain of 2.3 kg/decade necessary to maintain fatfree mass suggested by Forbes. 27 Body weight appears to be the prime determinant of fat-free mass changes in young and old. As previously mentioned, Williams et al. 25 noted in their cross-sectional study that weight gain during adulthood persisted in spite of considerable physical activity. They suggested that exerciseinduced weight loss offsets weight gain during middle age. Our results also suggest that physical activity minimized weight and BF, at least at the overall lower rates of weight gain seen in European than US populations 28 and that BF increased in sedentary adults. 23 Physical activity also minimized the decrease in FFMI with age. Clinical application of fat-free mass index and body fat mass index The use of height-normalized FFMI and BFMI further permits comparison of subjects with different heights. FFMI and BFMI each have a separate role in nutritional assessment, namely FFMI serves to determine whether or not fat-free mass is adequate and BFMI to evaluate overweight and obesity. While the FFMI or BFMI in healthy, normally active subjects often fall within the same category as their BMI category, this is not true in ill or inactive subjects. These subjects may have a normal BMI, but low FFMI and/or high BFMI. We have data that show that low FFMI is associated with increased length of stay. 14 Thus the use of FFMI and BFMI in studies of health risk could aid in determining the effects of too little muscle mass or too much fat. Classifying subjects by FFMI or BFMI may not make the nutritional assessment easier or quicker. However, classifying subjects into low, normal or high is more precise in terms of actual mass, rather than relative mass, which is the case when using % BF. Just as a BMI has become the reference for judging subjects as under-, normal and overweight, we believe that the classifications of low, normal and high FFMI and BFMI can provide valuable information on the body compartments of individuals. Future studies that include body composition measurements should evaluate the relationships of fat-free and BF mass and disease prevention and mortality. Study limitations Menopausal status or hormone replacement therapy of our 445 years old women are not available. These factors have been reported to affect body composition. 9 It was assumed that there was no difference in menopausal status and hormone replacement therapy between sedentary and physically active women. Furthermore, smoking and dietary factors, which are likely to affect body composition, were not controlled for in this study. Our volunteer subjects were not randomly selected and the results may, therefore, not apply to the general population. However, our volunteers are fairly representative of the Geneva population in terms of median BMI. The median BMI was 23.7 for men and 21.9 kg/m 2 for women in this study versus a median BMI of 25.3 and 23.0 kg/m 2, respectively, in the randomly selected population, aged years in Geneva. 29 The BMI would be expected to be lower in the present study because 49% of men and women were o40 years, a greater proportion than in the above study. Random selection would not have guaranteed the recruitment of a physically active group. Overweight and obese subjects are somewhat underrepresented in this study, because of inclusion of only apparently healthy subjects. This would have resulted in an underestimation of differences in BFMI between sedentary and physically active subjects. Misclassifications of the subjects in the sedentary group are possible. When physical activity profile was unclear, subjects were classified as sedentary. Misclassification might have caused some of the physically active subjects to be put in the sedentary group. This could have caused the mean BFMI to be lower in the sedentary group and the BFMI differences between the two groups might to be slightly underestimated. The misclassification would not have affected the conclusions that the sedentary subjects have larger BF than the physically active subjects. The effects of physical activity on body composition in elderly subjects must be interpreted with caution due to the small number of physically

98 ARTICLE IN PRESS Aging, physical activity and body composition 87 active subjects over 65 years, and no physically active women over 75 years being included. BIA measurements were not restricted to 46h from food intake, because a decrease of o4 11 O or 3% of resistance 30 due to meal intake would influence fat-free mass only by 1.6% and FFMI or BFMI by o0.01%. Such a change has no effect on our data. Conclusion Physically active subjects are less likely to have low or high FFMI, and high or very high BFMI, and are more likely to have low BFMI. In contrast to common claim that fat-free mass decreases with age, we found that FFMI was stable until 74 years, but BFMI was higher in older compared to younger subjects. The use of FFMI and BFMI permits comparison of subjects with different height and age. Future studies should be directed towards the evaluation of low or high FFMI and BFMI as risk factors for chronic diseases. Acknowledgements We thank the Foundation Nutrition 2000 Plus for its financial support. Appendix A Fat-free mass index (FFMI) and body fat mass index (BFMI) values for corresponding body mass index (BMI) in healthy adults are given in Table 5. Table 5 BMI (kg/m 2 ) Categories of FFMI or BFMI FFMI (kg/m 2 ) BFMI (kg/m 2 ) Men Z30 Very high NA a Z High Z Normal r18.4 Low r16.6 r1.7 Women Z30 Very high NA a Z High Z Normal r18.4 Low r14.5 r3.8 a Not applicable: very high fat-free mass index does not indicate increased risk. References 1. Lee CD, Blair SN, Jackson AS. Cardiorespiratory fitness, body composition, and all-cause and cardiovascular disease mortality in men. Am J Clin Nutr 1999;69: Allison DB, Gallagher D, Heo M, Pi-Sunyer FX, Heymsfield SB. Body mass index and all-cause mortality among people age 70 and over: the longitudinal study of aging. Int J Obes 1997;21: Rissanen AM, Heli.ovaara M, Knekt P, Reunanen A, Aromaa A. Determinants of weight gain and overweight in adult Finns. Eur J Clin Nutr 1991;45: Pate RR, Pratt M, Blair SN, et al. Physical activity and public health: a recommendation from the centers for disease control and the American college of sports medicine. JAMA 1995;273: Blair SN, Kohl HW, Barlow CD, et al. Changes in physical fitness and all-cause mortality: a prospective study of healthy and unhealthy men. JAMA 1995;273: Thune I, Njolstad I, Lochen ML, Forde OH. Physical activity improves the metabolic risk profiles in men and women. Arch Intern Med 1998;158: Kyle UG, Gremion G, Genton L, Slosman DO, Golay A, Pichard C. Physical activity and fat-free and fat mass as measured by bioelectrical impedance in 3853 adults. Med Sci Sports Exerc 2001;33: Roubenoff R, Dallal GE, Wilson PW. Predicting body fatness: the body mass index vs estimation by bioelectrical impedance. Am J Public Health 1995;85: Guo SS, Zeller C, Chumlea WC, Siervogel RM. Aging, body composition, and lifestyle: the Fels longitudinal study. Am J Clin Nutr 1999;70: Segal KR, Dunaif A, Gutin B, Albu J, Nyman A, Pi-Sunyer FX. Body composition, not body weight, is related to cardiovascular disease risk factors and sex hormone levels in men. J Clin Invest 1987;84: Broadwin J, Goodman-Gruen D, Slymen D. Ability of fat and fat-free mass percentages to predict functional disability in older men and women. J Am Geriatr Soc 2001;49: Heitmann BL, Erikson H, Ellsinger BM, Mikkelsen KL, Larsson B. Mortality associated with body fat, fat-free mass and body mass index among 60-year-old Swedish men-a 22-year follow-up. The study of men born in Int J Obesity Rel Metab Disord 2000;24: Kyle U, Unger P, Dupertuis Y, Karsegard L, Genton L, Pichard C. Body composition in 995 acutely ill or chronically ill patients at hospital admission: a controlled population study. J Am Diet Assoc 2002;102: Kyle UG, Pirlich M, Schuetz T, Lochs H, Pichard C. Low fatfree mass index and very high fat mass index at hospital admission are associated with increased length of stay. Clin Nutr 2002;21: Taylor HL, Jacobs Jr DR, Schucker B, Knudsen J, Leon A S, Debacker G. A questionnaire for the assessment of leisure time physical activities. J Chron Dis 1978;31: Lukaski HC, Bolonchuk WW, Hall CB, Siders WA. Validation of tetrapolar bioelectrical impedance measurements to assess human body composition. J Appl Physiol 1986;60: Houtkooper LB, Lohman TG, Going SB, Howell WH. Why bioelectrical impedance analysis should be used for estimating adposity. Am J Clin Nutr 1996;64:436S 48S. 18. Kyle UG, Genton L, Karsegard L, Slosman DO, Pichard C. Single prediction equation for bioelectrical impedance analysis in adults aged yrs. Nutrition 2001;17:

99 ARTICLE IN PRESS 88 U.G. Kyle et al. 19. Kyle UG, Schutz Y, Dupertuis YM, Pichard C. Body composition interpretation: contribution of fat-free mass index and body fat mass index. Nutrition 2003;19: Krotkiewski M, Grimby G, Holm G, Szczepanik J. Increased muscle dynamic endurance associated with weight reduction on a very-low-calorie diet. Am J Clin Nutr 1990;51: Williamson DF, Kahn HS, Remington PL, Anda RF. The 10-year incidence of overweight and major weight gain in US adults. Arch Intern Med 1990;150: Kahn HS, Tatham LM, Rodriguez C, Calle EE, Thun MJ, Heath Jr CW. Stable behaviors associated with adults 10-year change in body mass index and likelihood of gain at the waist. Am J Pub Health 1997;87: Haapanen N, Miilunpalo S, Pasanen M, Oja P, Vuori I. Association between leisure time physical activity and 10- year body mass change among working-aged men and women. Int J Obes Relat Metab Disord 1997;21: Voorrips LE, van Staveren WA, Hautvast JGAJ. Are physically active elderly women in a better nutritional condition than their sedentary peers? Eur J Clin Nutr 1991; 45: Williams PT. Evidence for the incompatability of age-neutral overweight, age-neutral physical activity standards for runners. Am J Clin Nutr 1997;65: Forbes GB, Gallup J, Hursh JB. Estimation of total body fat from potassium-40 content. Science 1961;133: Forbes GB. Exercise, lean weight: the influence of body weight. Nutr Rev 1992;50: Flegal KM, Carroll MD, Kuczarski RJ, Johnson CL. Overweight and obesity in the United States: prevalence and trends, l960 l994. Int J Obes 1998;22: Beer-Borst S, Morabia A, Hercberg S, et al. Obesity and other health determinants across Europe: the Euralim Project. J Epidemiol Pub Health 2000;54: Kushner RF, De Vries PMJM, Gudivaka R. Use of bioelectrical impedance analysis measurements in the clinical management of patients undergoing dialysis. Am J Clin Nutr 1996;64:503S 9S.

100 APPENDIX 7

101 Total Body Mass, Fat Mass, Fat-Free Mass, and Skeletal Muscle in Older People: Cross-Sectional Differences in 60-Year-Old Persons Ursula G. Kyle, MS, RD,* Laurence Genton, MD,* Didier Hans, Veronique L. Karsegard,* Jean-Pierre Michel, MD, Daniel O. Slosman, MD, and Claude Pichard, MD, PhD* OBJECTIVES: To evaluate body composition parameters, including fat-free mass (FFM), appendicular skeletal muscle mass (ASMM), relative skeletal muscle mass (RSM) index, body cell mass (BCM), BCM index, total body potassium (TBK), fat mass, percentage fat mass (FM), and their differences between age groups and to evaluate the frequency of sarcopenia in healthy older subjects DESIGN: Cross-sectional, nonrandomized study. SETTING: Outpatient clinic. PARTICIPANTS: Ninety-one healthy men and 100 healthy women age 60 and older. MEASUREMENTS: FFM, ASMM, FM, and percentage fat mass by whole-body dual-energy x-ray absorptiometry; TBK, BCM, and TBK/FFM ratio by whole body potassium-40 counter. RESULTS: All lean body mass parameters were significantly (P.05) lower in subjects age 80 and older than in those age 70 to 79, except ASMM in women. Mean FFM was 4.2 kg (7.3%) lower in men age 80 and older than in those younger than 70 and 2.9 kg (6.8%) lower in women age 80 and older than in those younger than 70. The skeletal muscle mass, reflected by ASMM, decreased more than FFM. This suggests that nonskeletal muscle mass is proportionally preserved during aging. Forty-five percent of men and 30% of women were sarcopenic by definition of BCM index and 11.0% of men and women by definition of RSM index. CONCLUSIONS: Significant age-related differences exist in body composition of older men and women between age 60 and 95. The greater decrease in TBK and BCM than the decrease in FFM and skeletal muscle mass suggests changing composition of FFM with age. Lack of From the Departments of *Clinical Nutrition, Geriatrics, and Nuclear Medicine, Geneva University Hospital, Geneva, Switzerland. This study was supported by the Foundation Nutrition 2000Plus. Address correspondence to Claude Pichard, MD, PhD, Head, Clinical Nutrition and Diettherapy, Geneva University Hospital, 1211 Geneva, Switzerland. agreement between two independent sarcopenia indexes suggests that further refinement in the definition of a sarcopenia index is necessary. J Am Geriatr Soc 49: , Key words: body composition; fat-free mass; fat mass; skeletal muscle mass; dual-energy x-ray absorptiometry; total body potassium; sarcopenia; older Significant changes in body composition are known to occur with aging and are a consequence of imbalances between energy intake and needs associated an with increasingly sedentary lifestyle, which results in a progressive increase in fat mass (FM) and a progressive reduction in fat-free mass (FFM) with aging. The age-related loss of muscle mass and muscle strength or sarcopenia is prevalent in older people and is associated with disability, independent of morbidity. 1 However, the amount of muscle mass that delineates the degree and the prevalence of sarcopenia have not been determined. Recent advances in body composition measurements now allow the simultaneous measuring of fat and lean components and the determination of regional quantities of these components. Dual-energy X-ray absorptiometry (DXA) is used to explore changes in total and regional fat and muscle mass, including appendicular (leg and arm) skeletal muscle mass (ASMM). In addition, whole body counting of potassium (TBK) permits the evaluation of the quantitative relationship between body cell mass (BCM) and FFM 2 and their evolution. Although a number of studies 2 5 have evaluated the associations between body composition parameters and age in older subjects, none have reported FFM, ASMM, FM, BCM, and TBK, and their indices and physical activity levels in healthy subjects age 60 and older and compared their differences between age groups. For nutritional assessment purposes, Baumgartner 6 derived a relative skeletal muscle (RSM) index and defined sarcopenia as values 2 standard deviations (SDs) be- JAGS 49: , by the American Geriatrics Society /01/$15.00

102 1634 KYLE ET AL. DECEMBER 2001 VOL. 49, NO. 12 JAGS low the sex-specific mean for RSM index in healthy control subjects younger than 35. This index follows the approach taken to define body mass index (BMI) and normalizes for shorter stature of older subjects. TBK reflects the intracellular mass or BCM, as suggested by Moore et al. 7 BCM is the metabolically, oxygenconsuming portion of the FFM. TBK is the single best predictor of nutritional status because it represents BCM, the working portion of FFM. 4 TBK and BCM decrease with age, but do not maintain a constant relationship to each other across heterogeneous populations. 2 Thus low BCM or TBK in older subjects would also indicate deficient muscle mass or sarcopenia in aging. Therefore, we derived a BCM index (BCM (kg) divided by height squared (m 2 )) and defined sarcopenia as values 2 SD below the sexspecific mean for BCM index in healthy control subjects younger than The purpose of this study was to determine (1) body composition parameters, including FFM, ASMM, BCM, and TBK and their differences between age groups in healthy, physically active subjects age 60 to 94; (2) whether percentage changes for various FFM parameters differedbetween age groups; and (3) the prevalence of sarcopenia as determined by RSM and BCM indexes in 191 older study subjects. SUBJECTS AND METHODS Subjects One hundred ninety-one healthy ambulatory Caucasians (91 men and 100 women) age 64 to 94 (Table 1) were included in this study. Subjects were recruited through advertisement in local newspapers and invitations sent to members of leisure clubs for older people. Although subjects were nonrandomly selected, statistical analysis revealed no differences for height, weight, and BMI between subjects in this study and 151 age-matched older healthy men and 234 women (data not shown) in the Geneva area. Exclusion criteria were acute medical treatment; hospitalization within 3 months of measurement; involuntary weight loss of 2 kg in the last 3 months; physical handicap that might interfere with body composition measurement (amputation, paralysis, etc.), or inability to walk 300 m. All subjects signed an informed consent statement. The subjects were interviewed to obtain information on education, profession, smoking, alcohol consumption, self-reported medical history, and current medications. Incidence of heart disease, hypertension, hyperlipidemia, and diabetes mellitus were based on treatment by medications, not on laboratory parameters or physical examination. Incidence of arthritis was based on medication and self-reported pain. Women were also asked about age of menopause, number of children, and hormone replacement therapy. The Geneva University Hospital Ethics Committee approved the study protocol. Physical Activity Physical activity was determined using a frequency questionnaire. 9 The time spent on total physical activity (walking, bicycling, gardening, sports, and household activities) calculated by multiplying the frequency and duration of each activity in the previous week, summing the values across activities, and dividing by seven. In Switzerland, walking is a regular activity used predominantly for transportation purposes (e.g., shopping) and was therefore not considered as sports activity. The mean intensity of total activity and sports activity was calculated using the following intensity scores: 10 light housework, 0.703; walking and gardening, 0.890; heavy housekeeping, 1.368; and bicycling, Some example of intensity scores for sports activities are for fishing; for bowling; for walking, gymnastics for older people, and golf; and for running, cycling, tennis, swimming, and soccer. The intensity scores are based on net energy cost of activities in kj/hour. 10,11 Body Composition Measurements Body height was measured to the nearest 0.5 cm and body weight to the nearest 0.1 kg on a balance beam scale. Height and weight of both groups were normally distributed. Three women age 65 and older had BMIs of 18.0 kg/m 2. Medical history revealed no recent weight loss or illness. Thus they were considered healthy. Fat-Free Mass FFM and ASMM were determined using whole-body dual energy x-ray absorptiometry (DXA) (Hologic QDR-4500 instrument, Enhanced Whole Body 8.26a software version; Hologic Inc. Waltham, MA) In this scanning technique, an x-ray generator emits alternating pulsed radiation of two energies, 100 and 140 kilovolts (peak) (kv(p)), in a fan-beam mode. As they pass through the body, these two x-ray beams are attenuated because of the absorption and scattering of the photons. The attenuation is measured for every pixel of the body surface in a linear array of 216 detectors. A complex development measurement allows determination of bone mineral and soft tissue densities. Soft tissue can further be partitioned into FM and FFM, because they have different attenuation characteristics. The scanner was calibrated for bone mineral content with a rotating drum and for fat with an external Lucite-aluminum phantom. 16 The precision of the measurements is 1.0% and 2.0% for the FFM and FM, respectively, 14,17 which is consistent with or better than reported in the literature. The effective total body radiation dose is 5.2 msv. 16,18 Percentage of FM was measured using the manufacturer s default definition. ASMM was measured as the sum of the lean soft-tissue masses for the arms and legs as described by Heymsfield et al. 12 Visser et al. 15 validated the Hologic QDR-4500 instrument in older subjects and found that FFM was positively associated with FFM by four-compartment model (r , standard error of estimate 1.6 kg) and with computed tomography (CT) at all four leg regions (r ). FFM by DXA was higher ( kg) than FFM by four-compartment model ( kg; P.001). Visser et al. suggested that the value of 1.82 bone mineral content used in the DXA equation possibly underestimated bone mass, which may partly explain the overestimation of DXA muscle mass vs CT muscle mass observed in their study. They concluded that the results suggest that fan-beam DXA offers considerable promise for the measurement of FFM and leg muscle mass in older persons.

103 JAGS DECEMBER 2001 VOL. 49, NO. 12 FAT-FREE MASS, SKELETAL MUSCLE, AND FAT MASS AFTER AGE Body Cell Mass The potassium-40 body content was measured by using a whole-body scintillation counter. Subjects were reclined in a tilting chair and were placed in the field of view of a large sodium iodide crystal (203-mm diameter). The gamma photons emitted by the subject interacted with the crystal, sending an output signal to the spectrum analyzer. To minimized background interference, the subject and the detector were placed in a special room surrounded by thick steel walls. 19 The natural potassium-40 isotope exists at a known and constant natural abundance of %. It was measured by counting the total pulses recorded in the channels of the photo peak of this isotope for 30 minutes. The calculated TBK content represents the BCM from which the lean body mass (LBM) (LBMTBK) can be extrapolated. This technique in humans, at our institution, has an accuracy of 5% and a precision of 2%. 19 Others reported a variability of potassium-40 counting in an anthropometric phantom 5% and in humans around 5%. 20 BCM was calculated from TBK using the equation, BCM (kg) TBK (mmol). 21 This equation is based on two assumptions: (1) the average intracellular potassium content is 3 mmol/g nitrogen and (2) the nitrogen content is 0.04 g/g wet tissue. There is good support for these assumptions in weight-stable healthy men and women across a wide age spectrum. 22 Cohn et al. 22 found that intracellular potassium concentration was similar in men and women and was not influenced by age. Definition of Sarcopenia Subjects were classified as sarcopenic using two different indexes of RSM or BCM/body mass. Relative Skeletal Muscle Index The RSM index was derived as muscle mass divided by stature squared (kg/m 2 ). 1 The square of the height in the denominator of the RSM index effectively eliminates difference in ASMM associated with greater height in younger adults and with gender. 1 Baumgartner et al., in an older New Mexico population, reported RSM index as 7.26 kg/m 2 in men and 5.45 kg/m 2 in women. 6 Cutoff values 2 SD used in this study were 7.06 kg/m 2 in men and 5.53 kg/m 2 in women, 8 based on our healthy subjects age 18 to 54 (n 224). Body Cell Mass Index TBK is an indicator of BCM. The BCM index was derived as BCM divided by stature squared (kg/m 2 ). Sarcopenia by BCM index was defined as values of 2 SD below the sexspecific mean for BCM in a healthy, younger person (age 18 54, n 224), which was 8.83 kg/m 2 in men and 6.96 kg/m 2 in women. 8 These cutoff values were used in this study. Statistics Descriptive statistics were calculated for height, weight, percentage of ideal body weight (IBW), BMI, FFM, ASMM, FM, percentage fat mass, TBK, and TBK/FFM and are expressed as mean SD and ranges. Analysis of variance (ANOVA) was used to test for differences between age groups and differences between sarcopenic and normal muscle mass groups. Differences for body composition parameters for each age group compared with healthy young subjects (age 18 34) were calculated as: percentage difference parameter mean value of young/ mean value of young /mean height of young 170 cm if male and 160 cm if female 100. Statistical significance was set at P.05 for all tests. RESULTS Table 1 shows the anthropometric characteristics of the healthy older subjects. Height and weight were progressively lower in older men. BMI and percentage of IBW were highest in men age 60 to 69 and were nonsignificantly lower thereafter. Height was significantly lower in older women. Weight, BMI, and percentage of IBW were highest in women age 70 to 79 and lowest in women age 80 and older. Demographic Health Information Table 2 shows social, educational, and health parameters of healthy subjects. Weight, FFM, and percentage of FM were highest in least-educated women and most-educated men (data not shown, ANOVA nonsignificant). Twenty percent of men and 58% of women never smoked. Less than 10% of men and 12% of women smoked at the time of the study. Weight, FFM, and percentage of FM were highest in former smokers and lowest in current smokers in both sexes (data not shown, ANOVA nonsignificant). Highest incidence of chronic diseases was hypertension, Table 1. Anthropometric Characteristics of a Healthy Older Caucasian Population Characteristic years years years 80 years Men, n Height, mean SD (range), cm ( ) Weight, mean SD (range), kg ( ) BMI, mean SD (range), kg/m ( ) Percentage of IBW, mean SD (range), % (84 141) Women, n Height, mean SD (range), cm ( ) * Weight, mean SD (range), kg ( ) BMI, mean SD (range), kg/m ( ) Percentage of IBW, mean SD (range), % (76 151) Analysis of variance comparison between age groups: vs years *P.05; vs 80 years P.05. SD standard deviation; BMI body mass index; IBW ideal body weight (Metropolitan Life Insurance, 1983).

104 1636 KYLE ET AL. DECEMBER 2001 VOL. 49, NO. 12 JAGS Table 2. Education, Smoking Habits, Alcohol Consumption, Medications and Menopause/Childbearing History of Men and Women Characteristic Men n (%) Women n (%) Education than high school 16 (17.6) 43 (43.0) High school & apprenticeship 47 (51.6) 38 (38.0) College education 28 (30.8) 19 (19.0) Smoking Never 19 (20.9) 58 (58.0) Former 63 (69.2) 29 (29.0) Current 9 (9.9) 13 (13.0) Alcohol* 80 (79.1) 80 (80.0) Medications Arthritis 13 (14.3) 20 (20) Heart disease 19 (20.9) 16 (16.0) Hypertension 43 (47.3) 42 (42.0) Hyperlipidemia 12 (13.2) 8 (8.0) Diabetes 6 (6.6) 4 (4.0) Thyroid 3 (3.3) 4 (4.0) Bone disease 18 (18.0) Abnormal bone scan (DXA) 50 (54.9) 62 (62.0) Time since menopause, yrs *Mean alcohol consumption: men g/day, women g/day See results DXA dual energy x-ray absorptiometry. with 47% of men and 42% of women reporting taking medications. Incidence of diabetes mellitus with treatment was low (6.6% in men and 4% in women). Over 50% of subjects had abnormal DXA bone scans showing either osteopenia (T-score 1.00) or osteoporosis (T-score 2.50) in femoral neck or lumbar region (L2 4) or both. Eighteen percent of women were being treated for osteoporosis. Effects of Aging on Lean Tissues Table 3 and Figure 1 show the differences in body composition parameters with age. All LBM parameters were progressively lower in older subjects. Overall, FFM was 4.2 kg lower in men age 80 and older than in those age 60 to 69 and 2.9 kg lower in women age 80 and older than in those age 60 to 69. Furthermore, BCM was 13.9% lower in men and 16.0% lower in women age 80 and older than in those age 60 to 69. Figure 1 shows that the differences in mean BCM and BCM index, and FFM in men but not women, between subjects age 60 to 69 and those age 80 and older were greater than the differences in ASMM and RSM index, suggesting greater depletion in BCM than ASMM with age. FM and percentage of fat mass were higher in subjects age 70 to 79 than in those age 60 to 69 or 80 and older. Although physical activity levels decreased with age, the decline was not significant. Percentage of Difference in Body Composition Parameters Table 3 also shows the percentage difference in FFM, ASMM, BCM, TBK, and TBK/FFM compared with values in healthy men and women age 18 to 34, normalized for lower height in older subjects. The percentage difference in ASMM was greater than the percentage difference in FFM, which this suggests that ASMM decline was proportionally greater than the loss of organ muscle mass. The percentage changes in TBK and BCM were approximately twice the decrease in FFM and skeletal muscle mass. Thus BCM the metabolic active mass decline was greater than either FFM or ASMM. This is confirmed by the significantly lower BCM/FFM ratio (0.46 in men age 80 and older vs 0.49 in those age 60 to 69, P.003; 0.44 in women age 80 and older vs 0.49 in those age 60 to 69, P.001, data not shown) and the lower TBK/FFM ratio. This would suggest that FFM of older subjects no longer has the same composition or TBK content as the FFM of younger subjects. The percentage difference in FM was greatest in subjects age 70 to 79 and was greater in men than in women. Physical Activity Forty-six percent of men and 41% of women regularly participated in sports activity (data not shown). Forty-two percent of men age 80 and older participated in sports activity, with a mean of 41 minutes/day, versus 64% men age 60 to 69, with a mean of 96 minutes/day. In women, neither participation in sports activity nor duration (mean 45 minutes/day) decreased with age. Total physical activity, which included household, walking, and sports activities (minutes/day), was positively associated with BCM (men r 0.299, P.004; women r 0.224, P.02), ASMM (men r.238, P.02, women ns), and TBK/FFM (women r 0.262, P.008, men ns) but not with FFM, FFM index, or RSM index (data not shown). Percentage of FM was negatively associated with total physical activity in women only (r 0.209, P.04). In men, sports activity (minutes/week) was positively associated only with BCM (r 0.46, P.0008), ASMM (r 0.24, P.002), and TBK/FFM (r 0.301, P.04). Sarcopenia of Aging Table 4 shows the differences in body composition parameters using the sarcopenia indexes: RSM and BCM index. Ten (11.0%) men and 11 (11.0%) women were classified as sarcopenic by definition of RSM index. Alternatively, by definition of BCM index, 41 (45.1%) men and 30 (30.0%) women were considered sarcopenic. Relative Skeletal Muscle Index By definition of RSM index, all lean body mass parameters except TBK/FFM ratio were significantly lower in men and women classified as sarcopenic than classified as having normal muscle mass. Percentage of FM was significantly lower in sarcopenic women compared with normal muscle mass women and nonsignificantly lower in sarcopenic men. Total physical activity was significantly different only in women with normal muscle mass compared with sarcopenic women. Correlations between RSM index and total physical activity were nonsignificant (r in men and r in women, data not shown).

105 JAGS DECEMBER 2001 VOL. 49, NO. 12 FAT-FREE MASS, SKELETAL MUSCLE, AND FAT MASS AFTER AGE Table 3. Body Composition Differences Between Age Groups in Healthy Adults Men Women Age yrs yrs yrs 80 yrs yrs yrs yrs 80 yrs n Fat-free mass, kg a, % ASMM, kg a, % RSM index, kg/m a, % Body cell mass, kg * a, % BCM index, kg * a, % TBK, g * a, % TBK/FFM, g/kg * a, % Fat mass, kg a, % % fat mass a, % Physical activity, min/d FFM fat-free mass, ASMM appendicular skeletal muscle mass, RSM relative skeletal muscle mass, BCM body cell mass, TBK total body potassium. a Percent difference compared to yrs old men and women, normalized for lower height in older subjects (170 cm for men, 160 cm for women). Analysis of variance comparison between age groups (adjusted for height and weight): vs y * p 0.05, p 0.001, vs 80 y p 0.05, p

106 1638 KYLE ET AL. DECEMBER 2001 VOL. 49, NO. 12 JAGS parameters were significantly lower in subjects age 80 and older than in those age 70 to 79, and prevalence of sarcopenia was higher in oldest healthy subjects. Furthermore, we found greater depletion with age of BCM than of FFM and ASMM, which suggests a FFM composition change. Figure 1. Fat-free mass (FFM) (top), appendicular skeletal muscle mass (ASMM), and body cell mass (BCM) (middle), normalized for height (men 170 cm, women 160 cm) (top) and relative skeletal muscle (RSM) index and BCM index (bottom) in men (left) and women (right) years, years, 80 years. The horizontal lines of the box plots represent the 25th, 50th (median), and 75th percentile and the error bars 10th and 90th percentiles. Body Cell Mass Index By definition of BCM index, all lean body mass parameters were significantly lower in men and women classified as sarcopenic than those classified as having normal muscle mass. Percentage of FM differed significantly between groups in men and women. Although total physical activity was higher in subjects with normal muscle mass than in sarcopenic men and women, the difference was not significant. DISCUSSION The age-related loss of muscle mass, or sarcopenia, is prevalent in older people and is strongly associated with disability, independent of morbidity, 1 and occurs even in healthy older persons. 3,5 We found that lean body mass Body Composition Changes with Age Our study confirms that older subjects have less FFM and ASMM mass than do younger subjects and that the decrease in ASMM with increasing age is greater in men than in women. Contrary to Gallagher et al., 5 who suggested a linear decrease of ASMM with age, our data suggest a greater decline in subjects age 80 and older than in those younger than 80 (Table 3). Furthermore, the greater percentage difference in ASMM than in FFM would suggest that the loss of ASMM is greater than the loss of organ muscle mass. This is confirmed by the lower RSM index, which indicates loss of muscle mass after controlling for body mass. These results agree with Cohn et al., 22 who reported a greater decrease with age in muscle than nonmuscle lean (organ) mass and TBK. The results of this study also indicate that BCM and FFM do not maintain a constant relationship to each other. The significantly greater differences with age in BCM and TBK than in FFM suggest that the composition of FFM is altered in older subjects. This is confirmed by the significantly lower BCM/FFM ratio. A greater decrease with age in TBK than FFM and greater decrease in the ratio of TBK/FFM in women than in men has been previously reported. 22 Lower TBK/FFM and BCM/FFM noted in older subjects suggest that the composition or quality of FFM is altered with age. Kehayias et al. 4 suggested that potassium depletion might be explained by a loss of BCM because of insufficient replacement of cells. Chumlea et al. 23 suggested that changes in body composition after age 70 may be a risk factor for subsequent health or function problems and may appear to herald the onset of, or the occurrence with, disease, poor health, and functional disability. Further research is necessary to determine whether the accelerated loss of ASMM and FFM in people age 80 and older is physiological or pathological. In view of the fact that our subjects were relatively active, we would anticipate that lower FFM, ASMM, and BCM and greater changes in these parameters would have been noted in less active subjects or subjects with chronic disease. The FFM, ASMM, and RSM index found in our older subjects at the levels reported did not appear to cause physical limitations because less than 20.9% of men and 15% of women reported fewer than 120 minutes/day of physical activity. Our results confirm previous observations 23 that the accumulation with age of FM in men and women occurs primarily before age 60 and that there is little or no further gain during old age. In addition, our data confirm previous reports 3 that FM, in terms of absolute FM as well as percentage of FM may be relatively stable in older men but may decrease with age in older women. Physical Activity We found nonsignificant lower physical activity levels in sarcopenic subjects than in subjects with normal muscle

107 JAGS DECEMBER 2001 VOL. 49, NO. 12 FAT-FREE MASS, SKELETAL MUSCLE, AND FAT MASS AFTER AGE Table 4. Body Composition and Physical Activity by Sarcopenic Classification in Men and Women Men Women Index Sarcopenic Normal Muscle Mass t-test Sarcopenic Normal Muscle Mass t-test RSM index n p p FFM, kg ASMM, kg RSM index, kg/m BCM, kg BCM index, kg/m TBK/FFM, g/kg Fat mass,% Total activity, min/d BCM index n p p FFM, kg ASMM, kg RSM index, kg/m BCM, kg BCM index, kg/m TBK/FFM, g/kg Fat mass, % Total activity, min/d FFM fat-free mass, ASMM appendicular skeletal muscle mass, RSM relative skeletal muscle mass, BCM body cell mass, TBK total body potassium. Analysis of variance comparison, significance level p mass. Although there was a significant association between physical activity and BCM in men and women, and ASMM and TBK/FFM in men only, cause and effect between physical activity and lean mass and FM cannot be determined. Low muscle mass has been associated with increased physical disability. 24 Alternatively, Visser et al. 25 found no association between low muscle mass and disabilities but suggested that a high percentage of FM was associated with higher levels of disability. Results might have been considerably different in more-obese subjects. Obesity appears to protect against loss of FFM in older people. 26 It is currently not known at which point excess FM offsets the advantages of higher FFM in terms of limiting mobility and affecting survival. Prevalence of Sarcopenia with Aging At present, there are insufficient data for forming any consensus about what constitutes deficient muscle mass or sarcopenia. Despite the fact that 45% of men and 30% of women were sarcopenic by definition of BCM index, we did not find decreased mobility in our subjects. We hypothesize that the absence of recent weight loss, the relatively small number of obese subjects (7.9% BMI 30 kg/m 2 ), and the low number of subjects with high FM (25% with 27% in men and 38% in women) and relatively active lifestyle accounted for this. Visser et al. 25 found that low skeletal muscle mass was not associated with self-reported physical disability and that persons with high percentages of body fat had high disability. However, our data did not include muscle strength measurements. It can therefore not be excluded that low muscle strength is present in subjects with low BCM index. Further research is necessary to determine the association between muscle strength and muscle fiber area and BCM index, and the level at which BCM depletion or sarcopenia affects muscle function. The lower prevalence (11%) of sarcopenia by definition of RSM index compared with BCM index in our healthy subjects suggests that the RSM index might be a better predictor of functional level in healthy older subjects because the functional level was not impaired in our subjects. Melton at al. 27 reported a prevalence of sarcopenia in 6% to 15% of subjects age 65 and older, depending on the muscle mass parameter that was evaluated. Higher rates of sarcopenia were reported in men than in women. 27 Melton et al. did not report BCM. Thus, the discrepancy in agreement of incidence between the two independent indices of sarcopenia suggests that further refinement in the definition of a sarcopenia index is necessary. Limitations of Study Because the data are cross-sectional, cause and effect cannot be determined. Factors other than aging may well be responsible for the changes reported. It is well known that improved food availability and public health has resulted in larger body size in younger subjects. We have partially compensated for this by normalizing the height of all subjects (170 cm in men, 160 cm in women) and by using RSM and BCM index, which effectively eliminated differences in ASMM and BCM associated with greater height in younger adults and with gender. The subjects in this study were volunteers in good health and may not be representative of the general older population. Only a small number of subjects in this study

108 1640 KYLE ET AL. DECEMBER 2001 VOL. 49, NO. 12 JAGS were underweight (BMI 19 kg/m 2, 2.6%) or obese (BMI 0 kg/m 2, 7.9%). Because we cannot definitively determine whether frail older subjects with low skeletal muscle mass died or were not included because of preexisting illness, the generalization of our findings is limited to healthy older people. Although the subjects in this study reported the presence of some chronic diseases (hypertension, heart disease, hyperlipidemia, and diabetes mellitus), the chronic diseases were not severe enough to limit physical activity, impede daily living activities, or cause overall mobility problems. Only about one-sixth of subjects reported arthritis, a condition that might reduce activity levels and lead to lower FFM and FM. The absence of mobility problems and the relative high prevalence of regular physical activity appear to have aided in limiting the loss of FFM and ASMM. Similarly, the results might not apply to subjects with greater prevalence of obesity, because obesity reduces loss of FFM in older people 26 and differences between age groups would likely be different. Comparisons with other studies are difficult because body composition parameters vary with age, weight, height, physical activity level, and general health status. CONCLUSION Skeletal muscle mass, reflected by ASMM, decreased more than FFM. This suggests that nonskeletal muscle mass is proportionally preserved during aging. The greater decrease in TBK and BCM than the decrease in FFM and skeletal muscle mass would suggest that the composition or quality of FFM is changing with age. Furthermore the prevalence of sarcopenia in healthy older supports the need for clinical assessment of lean tissue in older subjects. ACKNOWLEDGMENTS We are indebted to Giulio Conicella, Luc Terraneo, and Sophie Namy for technical assistance. REFERENCES 1. Baumgartner RN, Koehler KM, Gallagher D et al. Epidemiology of sarcopenia among the elderly in New Mexico. Am J Epidemiol 1998;147: Gallagher D, Visser M, Wang Z et al. Metabolically active component of fatfree body mass: Influences of age, adiposity and gender. Metabolism 1996; 45: Baumgartner RN, Stauber PM, McHugh D et al. Cross-sectional age differences in body composition in persons 60 years of age. J Gerontol A Biol Sci Med Sci 1995;50A:M307 M Kehayias JJ, Fiatarone MA, Zhuang H et al. Total body potassium and body fat: Relevance to aging. Am J Clin Nutr 1997;66: Gallagher D, Visser M, De Meersman RE et al. Appendicular skeletal muscle mass: Effects of age, gender and ethnicity. J Appl Physiol 1995;83: Baumgartner RN. Body composition in healthy aging. Ann NY Acad Sci 2000;904: Moore FD, Boyden CM. Body cell mass and limits of hydration of the fatfree body: Their relation to estimated skeletal weight. Ann NY Acad Sci 1963;110: Kyle UG, Genton L, Hans D et al. Age-related differences in fat-free mass, skeletal muscle, body cell mass and fat mass between yrs. Eur J Clin Nutr 2001;55: Bernstein MS, Morabia S, Sloutskis D. Definition and prevalence of sedentarism in an urban population. Am J Public Health 1999;89: Voorrips LE, Ravelli ACJ, Dongelmans PCA et al. A physical activity questionnaire for the elderly. Med Sci Sports Exerc 1991;23: Bink B, Bonier FH, Van der Sluys H. Assessment of the energy expenditure by indirect time and motion study. In: Edang K, Anderson KL, eds. Physical Activity in Health and Disease: Proceedings of the Bertostolen Symposium. Oslo: Oslo University, 1966, pp Heymsfield SB, Smith R, Aulet M et al. Appendicular skeletal muscle mass: Measurement by dual-photon absorptiometry. Am J Clin Nutr 1990;52: Kohrt WM. Preliminary evidence that DEXA provides an accurate assessment of body composition. J Appl Physiol 1998;84: Slosman DO, Casez JP, Pichard C et al. Assessment of whole-body composition using dual X-ray absorptiometry. Radiology 1992;185: Visser M, Fuerst T, Lang T et al. Validity of fan-beam dual-energy X-ray absorptiometry for measuring fat-free mass and leg muscle mass. J Appl Physiol 1999;87: Blake GM, Wahner HW, Fogelman I. The Evaluation of Osteoporosis: Dual Energy X-Ray Absorptiometry and Ultrasound in Clinical Practice. London: Martin Dunitz, Mazess RB, Peppler WW, Gibbons M. Total body composition by dual-photon (153 Gd) absorptiometry. Am J Clin Nutr 1990;40: Lewis MK, Blake GM, Fogelman I. Patient dose in dual x-ray absorptiometry. Osteoporosis Int 1994;4: Wenger P, Soucas L. Anthropogammamètre (whole body counter) de Genève. Etalonnage pour le potassium naturel et variation de la teneur en potassium naturel en fonction du poids du sujet. Helv Chim Acta 1964;47: Lukaski HC, Mendez J, Buskirk ER et al. A comparison of methods of assessment of body composition including neutron activation analysis of total body nitrogen. Metabolism 1981;30: Moore FD, Oleson KH, McMurrey JD et al. The Body Cell Mass and Its Supporting Environment. Philadelphia: W.B. Saunders, Cohn SH, Vaswani AN, Yasumura S et al. Assessment of cellular mass and lean body mass by noninvasive nuclear techniques. J Lab Clin Med 1985; 105: Chumlea WC, Vellas B, Guo SS. Malnutrition or healthy senescence. Proc Nutr Soc 1998;57: Hyatt RH, Whitelaw MN, Bhat A et al. Association of muscle strength with functional status of elderly people. Age Ageing 1990;19: Visser M, Harris TB, Langlois J et al. Body fat mass and skeletal muscle mass in relation to physical disability in very old men and women of the Framingham Heart Study. J Gerontol A Biol Sci Med Sci 1998;53A:M214 M Roubenoff R, Hughes VA. Sarcopenia: Current concepts. J Gerontol A Biol Sci Med Sc 2000;55:M716 M Melton III LJ, Khosla S, Crowson CS et al. Epidemiology of sarcopenia. J Am Geriatr Soc 2000;48:

109 APPENDIX 8

110 Clinical Nutrition xxx (2011) 1e7 Contents lists available at ScienceDirect Clinical Nutrition journal homepage: Original article Body composition changes over 9 years in healthy elderly subjects and impact of physical activity q Laurence Genton a, Véronique L. Karsegard a, Thierry Chevalley b, Michel P. Kossovsky c, Patrice Darmon d, Claude Pichard a, * a Clinical Nutrition, Switzerland b Rehabilitation and Geriatrics (TC), Switzerland c Primary Care Medicine (MP), University Hospital, Switzerland d Endocrinology and Nutrition Department (PD), Hôpital Nord, Marseille, France article info summary Article history: Received 15 October 2010 Accepted 13 January 2011 Keywords: Body composition Physical activity Older healthy subjects Background & aims: Age-related changes of body composition affect health status. This study aims at clarifying body composition changes in healthy elderly subjects, and evaluating the impact of physical activity on these changes. Methods: In 1999, 213 subjects 65 years recruited through advertisements underwent assessment of health state, energy expenditure by physical activity, body composition by bioimpedance analysis and body cell mass by total body potassium. In 2008, 112 of them repeated these assessments with additional determination of Barthel index, Mini Mental State Examination and Geriatric Depression Score. Results: Lean tissues decreased in both genders (p < 0.05). Compared to subjects aged 65e74 years at baseline, those aged 75 years lost more body weight (men: vs kg, women: vs kg, both p < 0.05), and fat-free mass (men: vs kg, women: vs kg, both p < 0.05). Plotting of fat-free mass evolution against age at baseline showed an exponential loss of fat-free mass. Increased physical activity limited lean tissue loss in men but not in women. Conclusion: Loss of lean tissues occurs exponentially with aging. Further research should confirm these changes in subjects over 80 years. Increasing physical activity limits fat-free mass loss in men but not women. Ó 2011 Elsevier Ltd and European Society for Clinical Nutrition and Metabolism. All rights reserved. 1. Introduction Body composition changes with aging. Our previous crosssectional study showed that in white subjects > 65 years, weight and fat-free mass (FFM) decreased, while fat mass (FM) tended to increase in men and stabilize or decrease in women. 1 Body cell mass (BCM) which is the metabolically, oxygen-consuming compartment of the FFM also decreases with age in cross-sectional studies, although to a lesser extent than FFM. 2 Most longitudinal studies confirmed the loss of FFM but could not detail at which age it occurred and showed conflicting results regarding FM. 3e5 q Conference presentation: Part of this article was presented at the ESPEN conference in Vienna, * Corresponding author. Rue Gabrielle Perret-Gentil 4, Geneva University Hospital, 1211 Geneva 14, Switzerland. Tel.: þ ; fax: þ address: claude.pichard@unige.ch (C. Pichard). However, clarification of body composition changes with aging is essential as they are related to health status and physical function. The decline of FFM with aging and the associated physical impairment, termed sarcopenia, results from loss of appendicular skeletal muscle mass (ASMM). 5 The features of sarcopenia are a loss of motor unit number, an atrophy of muscle fibers and a decline in protein synthesis. Its complex etiology encompasses decreased anabolic hormones, specific nutritional deficiencies (protein and vitamin D), changes of mitochondrial function of muscle cells, as well as increased pro-inflammatory cytokines, apoptotic activities in the myofibers and oxidative stress. 6 Whatever the origin of FFM loss, it is associated with a high risk of disability, balance disorders and falls, immune and muscular dysfunctions, and impaired quality of life. 7 In contrast, an increased FM is related to overweight and obesity, which in turn increases risk of cardio-vascular disease, certain type of cancers, osteoarthritis, liver and gallbladder diseases, sleep apnea and respiratory problems. Furthermore, gain in abdominal fat, which can be roughly evaluated by increased /$ e see front matter Ó 2011 Elsevier Ltd and European Society for Clinical Nutrition and Metabolism. All rights reserved. doi: /j.clnu Please cite this article in press as: Genton L, et al., Body composition changes over 9 years in healthy elderly subjects and impact of physical activity, Clinical Nutrition (2011), doi: /j.clnu

111 2 L. Genton et al. / Clinical Nutrition xxx (2011) 1e7 waist circumference or waist to hip ratio, increases the risk of functional limitations and disability. 8 Thus, neither the loss of FFM nor the gain of FM is beneficial for health. The fact that evolution of body composition with aging has not yet been clearly defined probably relies on confounding factors, as physical activity and diseases. In many studies, physical activity did not prevent loss of FFM or gain of total FM with aging 3,9 but frequency, type, intensity and duration of physical activity as well as length of follow-up may have been insufficient to lead to any effect. Furthermore, numerous diseases, such as malignancies, infections or decubitus ulcers are associated with decreased FFM. 10 This study aims at determining, in healthy elderly subjects at baseline and follow-up, 1) the 9-year longitudinal changes of anthropometrics and body composition measured by bioelectrical impedance analysis (BIA) and K 40 total body potassium (TBK), and 2) the influence of daily energy expenditure through physical activity on these changes. We have hypothesized that this prospective study will confirm our previously described crosssectional changes of body composition and that maintenance of energy expenditure through physical activity limits the loss of FFM. 2. Material and methods 2.1. Subjects In 1999, 213 healthy white subjects aged 65 years (108 women, 105 men) were recruited through advertisements in newspapers and leisure clubs, as part of a cross-sectional study which validated the Geneva BIA formula against dual-energy X-ray absorptiometry (DXA). 11 Exclusion criteria were acute heart, lung, kidney or liver failure, symptomatic neurological disorders, known active cancer or infectious diseases, significant mental impairment, involuntary weight loss and hospitalization in the last 6 months. After agreement of the volunteers, we sent to them an invitation letter, a health questionnaire and a physical activity questionnaire, asking the volunteers to answer the questions to the best of their knowledge. On the study day, the subjects presented themselves at the Geneva University Hospital between 7:30 and 9:00 am for one 2-h session. They reviewed the questionnaires with the investigators and underwent measurements of body composition by BIA, TBK and DXA. In 2008, we repeated the measurements of body composition by BIA and TBK as well as the health and physical activity questionnaires in all subjects included in the baseline study, who could be contacted by telephone, were able to participate and consented to do so. In addition, we performed an assessment of functional limitations by Barthel index, Mini Mental State Examination, Geriatric Depression Score (Fig. 1). All subjects signed an informed consent. The Geneva University Hospital Ethical Committee approved the protocol Anthropometric measurements Body height was measured to the nearest 0.5 cm with a height gauge and body weight to the nearest 0.1 kg on a balance beam scale (Seca, Germany), with the subject in underclothing and without shoes. Body mass index (BMI) was calculated as weight (kg)/height 2 (m). Waist circumference was measured as the part of the trunk located midway between the lower costal margin and the iliac crest while the subject was standing. Hip circumference was measured at the maximal circumference over the buttocks and allowed calculation of waist to hip ratio Bioelectrical impedance analysis The calculation of body composition by BIA relies on the geometrical relationship between impedance (Z), length (L) and Fig. 1. Two-hundred and thirteen subjects were included at baseline, 47% (n ¼ 101) were lost to follow-up. Among the subjects lost to follow-up, one fifth had died within the 9 years. volume (V) of an electrical conductor: V ¼ rl 2 /Z (r: specific resistivity). In the human body, V represents the volume of FFM, and L the height of the subject. Z is composed of the pure resistance R of the conductor, the FFM, and the reactance Xc produced by the capacitance of cellular membranes, tissue interfaces and non-ionic tissues. Practically, the skin was cleaned with 70% alcohol. Adhesive electrodes (3M Red Dot, *M Health Care, Borken, Germany) were placed on the right hand and foot, the subject lying on his back as generally described. A generator (Xitron, 4000B, Xitron Technologies, San Diego, Calif., USA) applied an alternating electrical current of 50 khz and 0.8 ma to these electrodes. Resistance and reactance were measured. FFM was calculated by the following equation, 12 validated against DXA during the baseline visit 11 : FFM (kg) ¼ þ (0.518 height 2 /R) þ (0.231 weight) þ (0.13 reactance) þ (4.229 sex), where sex ¼ 1 for men and 0 for women. Overall precision of FFM measured by BIA is 2.7e4.0% in healthy adults. 13 Appendicular skeletal muscle mass (ASMM) was calculated with our previously published equation, validated against DXA 14 : ASMM (kg) ¼ þ (0.267 height 2 /resistance) þ (0.095 weight) þ (0.012 age) þ (0.058 reactance) þ (1.909 sex) where sex ¼ 1 for men and 0 for women. The error for predicting ASMM was 1.1 kg or 5% in volunteers and 1.5 kg or 7.6% in patients. Please cite this article in press as: Genton L, et al., Body composition changes over 9 years in healthy elderly subjects and impact of physical activity, Clinical Nutrition (2011), doi: /j.clnu

112 L. Genton et al. / Clinical Nutrition xxx (2011) 1e Total body potassium TBK was measured with a whole body scintillation 40 K counter. The subjects sitting in a tilting chair were placed in the field of view of a large sodium iodide crystal (203 mm-diameter). The gamma photons emitted by the subjects interacted with the crystal and sent an output signal to the spectrum analyzer. To minimize background interference, the subject and the detector were placed in a room surrounded by thick steel walls. 40 K was measured by counting the total pulses recorded in the channels of the photo peak of this isotope for 30 min and divided by to obtain total body potassium (TBK) in millimoles. At our institution, this technique has an accuracy of 5% and a precision of 2%. TBK allows calculation of body cell mass (BCM), which is the working portion of lean body mass by the following formula: BCM (kg) ¼ TBK (mmol). 15 This formula relies on two assumptions: 1) the average intracellular potassium content is 3 mmol/g nitrogen and 2) the nitrogen content is 0.04 g/g wet tissue Physical activity Physical activity was reported with a validated frequency questionnaire and included leisure and non-leisure physical activities. 16 Energy expenditure by physical activity, defined as PA in the whole text, was calculated by multiplying the frequency and duration of each physical activity in the previous week, by the body weight and the metabolic equivalent (MET) of the analyzed activity. 17 Analyzed activities were all activities besides of sleeping. The energy expenditures of the different activities were added up and divided by 7 to obtain daily PA. We also calculated daily energy expenditure of moderate and vigorous activities, defined as 3 MET by Ainsworth et al., 17 using the same methodology Co-morbidity index and functional limitations The medical history allowed the determination of the Charlson co-morbidity index 18 This index encompasses 19 medical conditions weighted 1e6 with total scores ranging from 0 to 37. It is strongly associated with mortality and disability. 19 Functional limitations in 2008 were assessed by the Barthel index, where values of 75e100 indicate minimal or absence of neurologic disturbances, 20 the Mini Mental State Examination (MMSE), where values <20 indicate mild to severe cognitive impairment 21 and finally the 30-item Geriatric Depression Score (GDS), where values > 20 correspond to severe depression Statistics Results are shown as mean SD. Unpaired t-tests compared baseline anthropometrics and body composition between returnees and non returnees. Subjects were categorized in age classes 65e74 or 75 years according to age at baseline. These age classes were used because our normative body composition data were performed according to these age groups. 1 Paired t-tests compared baseline anthropometrics and body composition between 1999 and 2008 in subjects altogether or according to age class. Evolution of anthropometrics and body composition between age groups were compared by unpaired t-tests Multiple linear regressions evaluated the impact of baseline age, FFM, PA and PA changes on evolution of FFM, separately for each gender. Evolution of FFM according to age at baseline was represented graphically. PA was compared between 1999 and 2008 in subjects altogether or according to age class with paired t-tests. Evolution of PA between both age groups was compared with unpaired t-tests. Simple regressions evaluated the relationship between changes in PA and changes in body composition. Subsequently, subjects were categorized in two PA groups: those who increased and those who decreased PA between 1999 and Continuous variables were compared between both groups by unpaired t-tests. To evaluate the impact of physical activity intensity, among the subjects who increased PA, we compared subjects who increased moderate to vigorous PA (3 MET) with thosewho decreased PA by ManneWhitney test. Significance was set at p < Statistics were performed with Statview 5.0 (Abacus Concept, Berkeley, Calif., USA). 3. Results Of the 213 subjects included in 1999, 112 were measured again in Compared to non returnees, women returnees were younger ( vs years, p ¼ 0.01) and had higher PA ( vs k cal/d, p ¼ 0.08) but similar anthropometric characteristics and body composition. Men returnees were significantly younger ( vs years, p < 0.001), had higher FFM ( vs , p ¼ 0.03), ASMM ( vs , p ¼ 0.02) and PA ( vs , p < 0.001) than their non returnee counterparts. The Charlson co-morbidity index was missing in 2 women of the age class 65e74 years but reported in all men. It varied between 0 and 10 at baseline and follow-up. In 2008, subjects had a Barthel index between 75 and 100 ( ), an MMSE between 20 and 30 ( ) and a GDS between 0 and 26 (mean ). One men had a GDS of 26 but all others <20. These results confirmed that the included subjects were in a good health Body composition and aging In both genders, height, FFM, ASMM and BCM decreased significantly, while waist circumference and waist to hip ratio increased (Table 1). The changes of body composition according to age class are shown on Table 2. Evolution of weight and body composition differed between age classes (Fig. 2). The loss of weight, FFM and ASMM occurred in subjects 75 years but not in those aged 65e74 years. Regarding anthropometrics, waist circumference increased less in women 75 years than in those aged 65e74 years (p ¼ 0.048). BMI decreased more in women 75 years than in those aged 65e74 years (p ¼ 0.01). In men, height decreased more in the subjects aged 75 years than in those aged 65e74 years (p ¼ 0.01), while waist to hip ratio decreased less in subjects aged 75 years (p ¼ 0.03). The afore-mentioned results suggested a non-linear relationship of FFM with age at baseline, which could be demonstrated in Fig. 3. FFM loss occurred around 69 years in men and 72 years in women. In men, a low baseline FFM (p ¼ 0.03), a low baseline PA (p < 0.001) and a higher decrease of PA (p < 0.001) predicted a higher loss of FFM. This multivariate model explained 34% (adjusted R 2 ) of the variance of FFM loss. In women, the evolution of FFM was not affected by these variables Body composition and physical activity Physical activity was reported in 55 women (data missing in 2 women of the age class 65e74 years) and 54 men (data missing for one subject aged 75 years). In women, there was no significant decrease of PA between 1999 and 2008, whether considering women altogether ( vs k cal/d, p ¼ 0.82) or by age classes. In men, there was a significant decrease of PA between 1999 and 2008 ( vs k cal/d, p ¼ 0.01). Categorization in age classes showed a significant decrease of PA in men aged 65e74 years ( vs. Please cite this article in press as: Genton L, et al., Body composition changes over 9 years in healthy elderly subjects and impact of physical activity, Clinical Nutrition (2011), doi: /j.clnu

113 4 L. Genton et al. / Clinical Nutrition xxx (2011) 1e7 Table 1 Longitudinal measurements of body composition (mean SD). Women (n ¼ 57 a ) Men (n ¼ 55 b ) Δ2008e Δ2008e1999 Age (years) * * Height (cm) * * Weight (kg) Body mass index (kg/m 2 ) Waist circumference (cm) * * Waist/hip ratio * * Fat mass (kg) Fat free mass (kg) * * Fat free mass (%) Appendicular muscle mass (kg) * * Body cell mass (kg) * * * Significantly different from a Except for waist and hip circumferences where n ¼ 56 and body cell mass where n ¼ 36. b Except for body cell mass where n ¼ k cal/d, p ¼ 0.03) and a similar tendency in men aged 75 years ( vs k cal/d, p ¼ 0.06). Simple regression between changes in PA and changes in each anthropometric and body composition element were never significant. The classification of subjects in the 2 PA categories showed similar mean age and, as expected, significantly different PA changes between groups. PA changes were vs k cal/ d(p < 0.01) in women increasing or decreasing PA, respectively, and in men, PA changes were k cal/d vs k cal/ d(p < 0.01), respectively. Women of both PA groups had similar baseline characteristics (Table 3) as well as anthropometric and body composition changes (Fig. 4A). Baseline analysis of men showed that subjects increasing PA had a lower BMI, waist circumference and fat mass (Table 3). Increasing PA allowed preservation of body composition (Fig. 4B) and a higher gain in BMI ( vs kg/m 2, p ¼ 0.01) than decreasing PA. To evaluate the effect of physical activity intensity, we further compared, among the subjects who increased PA, the women who increased PA > 3MET(n ¼ 21) with those who did not (n ¼ 8). Women who increased PA 3 MET lost less weight ( vs kg, p ¼ 0.04), gained inwaist circumference ( cm vs , p ¼ 0.02) and waist to hip ratio ( vs , p ¼ 0.05), but could not prevent loss of FFM. In contrast, men who increased PA >3 MET(n ¼ 12) lost more BCM than those who decreased PA 3MET(n ¼ 7), with values of and kg respectively (p ¼ 0.04). The other changes in anthropometric and body composition were similar between both subgroups of men. Thus, in healthy elderly men and women, increase of PA limits or tends to limit the loss of FFM, ASMM and BCM. Intensity of physical activity may play a role in body composition changes, especially in women. 4. Discussion This 9-year longitudinal study showed that loss of body weight and lean tissues was higher in subjects aged 75 years at baseline compared to those aged 65e74 years. Plotting of age at baseline Table 2 Longitudinal measurements of body composition by age class (mean SD). Women Men e74 years n ¼ 36 n ¼ 39 Height (cm) * Weight (kg) * * Body mass index (kg/m 2 ) * Waist circumference (cm) * * Waist/hip ratio * * Fat mass (kg) Fat free mass (kg) Fat free mass (%) Appendicular muscle mass (kg) Body cell mass (kg) a * 75 years n ¼ 21 n ¼ 16 Height (cm) * Weight (kg) * * Body mass index (kg/m 2 ) Waist circumference (cm) b * Waist/hip ratio b * Fat mass (kg) Fat free mass (kg) * * Fat free mass (%) Appendicular muscle mass (kg) * * Body cell mass (kg) c * * Significantly different from 1999 (p < 0.05). a Women: n ¼ 21, men: n ¼ 27. b Women: n ¼ 20, n ¼ 16. c Women: n ¼ 13, men n ¼ 11. Please cite this article in press as: Genton L, et al., Body composition changes over 9 years in healthy elderly subjects and impact of physical activity, Clinical Nutrition (2011), doi: /j.clnu

114 L. Genton et al. / Clinical Nutrition xxx (2011) 1e7 5 A B Fig. 2. The changes of body composition in women (A) and men (B) are different whether the subjects were aged 65e74 years or 75 years., represent subjects aged 65e74 years (36 \, 39_) and - those aged 75 years (21 \, 16_). Men and women aged 75 years loose weight, FFM and ASSM compared to subjects aged 65e74 years. * indicates p < 0.05 between both age groups. against changes of FFM suggested the loss of FFM to occur around 70 years. Increased energy expenditure through physical activity limited the loss of lean tissues and was associated with increased weight and FM in men while it did not alter body composition in women. This study included subjects who were healthy at baseline and remained healthy after 9 years of follow-up, as reflected by their Charlson index below 10 at both visits. BMI was 25.1 and 26.1 kg/m 2 in women and men, respectively, at baseline and did not change significantly after 9 years. These ranges of BMI have recently been associated with the lowest 10-year mortality in subjects aged 70e75 years. 23 FFM (kg) at baseline was slightly higher than Fig. 3. Changes of FFM are plotted against age at baseline for women and men (mean SEE). The gray circles represent values for women and the black triangles the values for men. Regression curves are shown in gray for women (y ¼ 0.013x 2 þ 1.817x , r 2 ¼ 0.15, SEE ¼ 2.40) and black for men (y ¼ 0.027x 2 þ 3.707x , r 2 ¼ 0.19). Loss of FFM seems to start around 72 years in women and 69 years in men. published percentiles of healthy subjects of the same age 1 and was clinically not due to overhydration. Thus, our subjects did not present any sign of malnutrition, which itself is often associated with chronic diseases. Furthermore, the low prevalence of functional limitation in 2008 suggests that our subjects were not functionally disabled in 1999 and reinforces the statement that our subjects were and stayed healthy. Our study showed a significant loss of FFM, ASMM and BCM with aging in men altogether, despite of weight maintenance. Other longitudinal studies have also found a loss of FFM in older men but the amount and age of occurrence varies. 4,5,24e29 Hughes et al. included 131 subjects (61 8 years) and followed their body composition measured by hydrodensitometry for 9 years 24 The only exclusion criteria at follow-up were knee or hip arthroplasty. Men lost FFM ( kg over 9 years) to a similar extent then in our study. However, their subjects were younger and rather of the age class 65e74 years, in which we could not show any decrease of FFM. Whether these contradictions are related to different life styles, ethnicity, health states or methods of body composition measurements remains unclear. According to their exclusion criteria at followup, we can nevertheless assume their subjects to be less healthy than ours, which could explain the loss of FFM at younger age. In another American study, body composition changes were measured by DXA in 78 older subjects with a mean follow-up interval of 5 years. 5 Subjects were healthy according to medical history at follow-up. The authors assumed that loss of FFM with aging occurred linearly and expressed their results as annual rate of changes. They found a decrease of FFM ( kg/year) and ASMM in men ( kg/year) despite of weight maintenance. These losses seem higher than in our study despite of similar age range. However, follow-up interval varied between 1 and 9 years. Since we did not find a linear relationship between FFM loss and age at baseline (cf. Fig. 3), it is difficult to compare their results with ours. Five other longitudinal studies found comparable results to our study, but did not mention health state of included subjects at follow-up. One showed maintenance of FFM over 6.5 years in Scottish men with a mean age of 62 years at inclusion. 28 In our study, men of the age class 65e74 years did not decrease FFM either. The others described a loss of FFM in older subjects: kg FFM over 2 years in white subjects aged 74 3 years at baseline, kg over 3 years in Hong Kong Chinese men aged years at baseline, kg over 5 years in Swedes aged 75 years at baseline 27 and kg over 7 years in Italians aged 71 years at baseline 25 In our study, men 75 years showed a mean decrease of 3.6 kg FFM in 9 years in parallel with weight loss, thus a FFM loss similar to these studies. Regarding women altogether, they decreased FFM and ASMM but to a lesser extent than men. Three longitudinal study also showed a loss of FFM in women 25e27 but the others could not demonstrate it 4,5,24 Compared to the study of Hughes et al., 24 our subjects were older. When considering only women aged 65e74 years, we could not find any decrease in FFM either. The difficult comparison of our study with that of Gallagher et al. 5 has been explained before. Differential results between our study and that of Woo et al. 4 may rely on ethnic and environmental differences. One longitudinal study showed that the loss of FFM occurred already in younger subjects. 29 Forbes included 20 men and women, aged 24e53 years at baseline, and followed their body composition, measured by TBK, over 21e38 years. He described a progressive loss of 1.5 kg FFM per decade in subjects who maintained weight. However, the majority of his subjects had suffered from acute or chronic diseases during follow-up which may have decreased FFM, independently of aging, Furthermore, he had no information on PA and PA changes. The included subjects may have become more inactive with aging and thus have lost FFM. We have found that Please cite this article in press as: Genton L, et al., Body composition changes over 9 years in healthy elderly subjects and impact of physical activity, Clinical Nutrition (2011), doi: /j.clnu

115 6 L. Genton et al. / Clinical Nutrition xxx (2011) 1e7 Table 3 Baseline anthropometrics and body composition according to physical activity (PA) categories (mean SD). Women (n ¼ 55 a ) Men (n ¼ 54 b ) PA increasers (n ¼ 29) PA decreasers (n ¼ 26) PA increasers (n ¼ 19) PA decreasers (n ¼ 36) PA (kcal/d) * Age (years) Height (cm) Weight (kg) Body mass index (kg/m 2 ) * Waist circumference (cm) * Waist/hip ratio Fat mass (kg) * Fat free mass (kg) Fat free mass (%) Appendicular muscle mass (kg) Body cell mass (kg) * Significantly different between PA categories (p < 0.05). a n ¼ 55 except for waist circumference where n ¼ 54 (26 subjects increased and 28 decreased PA) and body cell mass where n ¼ 34 (15 women increased and 19 decreased PA). b n ¼ 54 except for body cell mass where n ¼ 38 (28 men increased and 10 decreased PA). FFM loss occurred only around 70 years. Forbes did not find an acceleration of FFM loss in subjects over 60 years but may not have included enough subjects reaching 70 years during follow-up. Thus, the apparently different results obtained by Forbes may be explained by different health states and PA changes in his study. The stability of FM in our aging subjects was in agreement with three longitudinal studies. 4,25,27 However, others showed an FM increase in one or both genders despite of stable weight. 5,24 Thus, longitudinal studies suggest that kinetic of FM in older subjects is variable and depends on other parameters than aging, as, for instance, evolution of health state, physical activity, medication, hormonal status or nutritional intakes. Although FM stays stable, fat distribution changes with aging. Indeed, waist to hip ratio increased in men and women overall and A B according to age class, and waist circumference increased in men and women aged 65e74 years, indicating increased abdominal fat. Severe functional limitation has been associated with waist circumferences 104 in men and 107 cm in women 30 and increased risk of cardio-vascular disease with values 102 cm for men and 88 cm for women. 31 Furthermore, values 103 cm in men and 89 cm in women, with BMI adjustment, doubled the rate of allcause mortality. 32 Mean values of our subjects were below these cut-offs at follow-up, suggesting again that they were in a good health. Ours results confirm the findings of Hughes et al., who showed that waist and waist to hip ratio of 129 subjects, aged 60 8 years at baseline increased significantly over 9 years. 33 One hypothesized way to counteract FFM is physical activity. Only one longitudinal study, beside ours, evaluated the impact of PA changes on body composition changes. 24 The authors reported weekly energy expenditure through sports and recreational activities during the past year collected through a 3-item questionnaire. Thus, by definition, they took into account less activities than we did and their baseline PA was clearly below that reported in our study. However, the trends between our studies were similar as we both noted a decrease in PA in aging men but not women. The decrease of PA in men may be partly linked to age-related hypogonadism and the subsequent decrease in FFM. 34 Hughes et al. did not find any relationship of PA changes with weight, FFM and FM changes, in men or women. We found the same results when using simple regressions. However, classification of subjects according to PA dynamics over 1999 and 2008 showed that men increasing PA maintained weight, FFM, ASMM and BCM. Furthermore, multiple regressions also demonstrated that PA changes adjusted for baseline PA affected evolution of FFM in men. In women, we cannot evaluate whether PA changes are related to a similar loss of FFM, as the PA changes during the 9 years of follow-up were not significant. Our results suggest that additional energy expended through activities other than leisure activities and sports may have a beneficial effect on body composition. The other previously mentioned longitudinal studies 4,5,28 did not evaluate changes in PA Study limitations Fig. 4. This figure shows the changes of body composition according to changes in physical activity energy expenditure (PA) (mean SEE). Results for women are shown on graph A and those for men on graph B. represent subjects who have decreased PA (26 \, 36_) and those who have increased PA (29 \, 19_) between 1999 and In men, increased PA allows preservation of FFM, ASMM, BCM as well as a slight increase in weight and FM. In women, results were similar, although not significant. * indicates p < 0.05 between both PA groups. Our study population was relatively small and subjects aged over 80 years at baseline were few. Thus, our findings on FFM changes according to age at baseline need further confirmation in subjects over 80 years. Body fat distribution was assessed indirectly through anthropometrics, and thus may not have been fully captured. We have no information on energy balance or dietary intake. Insufficient intakes may partly explain the weight and FFM Please cite this article in press as: Genton L, et al., Body composition changes over 9 years in healthy elderly subjects and impact of physical activity, Clinical Nutrition (2011), doi: /j.clnu

116 L. Genton et al. / Clinical Nutrition xxx (2011) 1e7 7 loss, especially in subjects 75 years. However, it is difficult to evaluate the dietary intakes over 9 years at only two time-points and the use of questionnaires often results in underreporting anyway. Furthermore, the physical activity questionnaire is subjective but should be able to evaluate PA changes of one subject. Finally, the metabolic equivalents assigned to physical activities may be overestimated in older subjects, as they generally cannot sustain similar activity intensity to younger subjects. 5. Conclusion This 9-year longitudinal study showed that loss of body weight and lean tissues occurs exponentially around 70 years in healthy older subjects. Increasing daily energy expenditure through physical activity limited the loss of lean tissue in men but not women. Further research should confirm these changes of body composition in subjects over 80 years. Statement of authorship We hereby certify that it is an original publication and the manuscript has not been previously submitted or published elsewhere. LG participated in the concept and design of the study, data acquisition, interpretation of data, and preparation of the manuscript. VLK participated in the concept and design of the study, data acquisition and interpretation of data. TC participated in interpretation of data and critical revising of the article. MPK participated in analysis of data and critical revising of the article. PD participated in data acquisition and interpretation of data. CP did the fundraising, participated in the concept and design of the study, interpretation of data and critical revising of the article. All authors approved the final version of the manuscript. Conflict of interest None of the authors has declared a conflict of interest. Acknowledgments We thank the Public Foundation Nutrition 2000Plus for financial support. We thank all the dietetic students who helped with the assessment of the questionnaires and the measurements of bioelectrical impedance analysis, Karine Jeandet and Sophie Namy for the measurements of total body potassium, as well as Didier Hans for his critical input. References 1. Kyle UG, Genton L, Slosman DO, Pichard C. Fat-free and fat mass percentiles in 5225 healthy subjects aged 15 to 98 years. Nutrition 2001;17:534e Kyle UG, Genton L, Hans D, Karsegard VL, Michel JP, Slosman DO, et al. Total body mass, fat mass, fat-free mass, and skeletal muscle in older people: crosssectional differences in 60-year-old persons. J Am Geriatr Soc 2001;49:1633e Raguso CA, Kyle U, Kossovsky MP, Roynette C, Paoloni-Giacobino A, Hans D, et al. A 3-year longitudinal study on body composition changes in the elderly: role of physical exercise. Clin Nutr 2006;25:573e Woo J, Ho SC, Sham A. Longitudinal changes in body mass index and body composition over 3 years and relationship to health outcomes in Hong Kong Chinese age 70 and older. J Am Geriatr Soc 2001;49:737e Gallagher D, Ruts E, Visser M, Heshka S, Baumgartner RN, Wang J, et al. Weight stability masks sarcopenia in elderly men and women. Am J Physiol Endocrinol Metab 2000;279:E366e Evans WJ. Skeletal muscle loss: cachexia, sarcopenia, and inactivity. Am J Clin Nutr 2010;91:1123Se7S. 7. Janssen I, Heymsfield SB, Ross R. Low relative skeletal muscle mass (sarcopenia) in older persons is associated with functional impairment and physical disability. J Am Geriatr Soc 2002;50:889e Houston DK, Stevens J, Cai J, Morey MC. Role of weight history on functional limitations and disability in late adulthood: the ARIC study. Obes Res 2005;13:1793e Carmeli E, Orbach P, Lowenthal DT, Merrick J, Coleman R. Long-term effects of activity status in the elderly on cardiorespiratory capacity, blood pressure, blood lipids, and body composition: a five-year follow-up study. Sci World J 2003;3:751e Roubenoff R. The pathophysiology of wasting in the elderly. J Nutr 1999;129:256Se9S. 11. Genton L, Karsegard VL, Kyle UG, Hans DB, Michel JP, Pichard C. Comparison of four bioelectrical impedance analysis formulas in healthy elderly subjects. Gerontology 2001;47:315e Kyle UG, Genton L, Karsegard L, Slosman DO, Pichard C. Single prediction equation for bioelectrical impedance analysis in adults aged 20e94 years. Nutrition 2001;17:248e Lukaski HC, Johnson PE, Bolonchuk WW, Lykken GI. Assessment of fat-free mass using bioelectrical impedance measurements of the human body. Am J Clin Nutr 1985;41:810e Kyle UG, Genton L, Hans D, Pichard C. Validation of a bioelectrical impedance analysis equation to predict appendicular skeletal muscle mass (ASMM). Clin Nutr 2003;22:537e Moore FD, Oleson KH, McMurrey JD, Parker JH, Ball M. The body cell mass and its supporting environment. Philadelphia: W.B. Saunders; Bernstein M, Sloutskis D, Kumanyika S, Sparti A, Schutz Y, Morabia A. Databased approach for developing a physical activity frequency questionnaire. Am J Epidemiol 1998;147:147e Ainsworth BE, Haskell WL, Whitt MC, Irwin ML, Swartz AM, Strath SJ, et al. Compendium of physical activities: an update of activity codes and MET intensities. Med Sci Sports Exerc 2000;32:S498e Hall WH, Ramachandran R, Narayan S, Jani AB, Vijayakumar S. An electronic application for rapidly calculating Charlson comorbidity score. BMC Cancer 2004;4: de Groot V, Beckerman H, Lankhorst GJ, Bouter LM. How to measure comorbidity. A critical review of available methods. J Clin Epidemiol 2003;56:221e Loewen SC, Anderson BA. Reliability of the modified motor assessment scale and the Barthel index. Phys Ther 1988;68:1077e Crum RM, Anthony JC, Bassett SS, Folstein MF. Population-based norms for the mini-mental state examination by age and educational level. JAMA 1993;269:2386e Yesavage JA, Brink TL, Rose TL, Lum O, Huang V, Adey M, et al. Development and validation of a geriatric depression screening scale: a preliminary report. J Psychiatr Res 1982;17:37e Flicker L, McCaul KA, Hankey GJ, Jamrozik K, Brown WJ, Byles JE, et al. Body mass index and survival in men and women aged 70 to 75. J Am Geriatr Soc 2010;58:234e Hughes VA, Frontera WR, Roubenoff R, Evans WJ, Singh MA. Longitudinal changes in body composition in older men and women: role of body weight change and physical activity. Am J Clin Nutr 2002;76:473e Rossi A, Fantin F, Di Francesco V, Guariento S, Giuliano K, Fontana G, et al. Body composition and pulmonary function in the elderly: a 7-year longitudinal study. Int J Obes (Lond) 2008;32:1423e Visser M, Pahor M, Tylavsky F, Kritchevsky SB, Cauley JA, Newman AB, et al. One- and two-year change in body composition as measured by DXA in a population-based cohort of older men and women. J Appl Physiol 2003;94:2368e Dey DK, Bosaeus I, Lissner L, Steen B. Changes in body composition and its relation to muscle strength in 75-year-old men and women: a 5-year prospective follow-up study of the NORA cohort in Goteborg, Sweden. Nutrition 2009;25:613e Murray LA, Reilly JJ, Choudhry M, Durnin JV. A longitudinal study of changes in body composition and basal metabolism in physically active elderly men. Eur J Appl Physiol Occup Physiol 1996;72:215e Forbes GB. Longitudinal changes in adult fat-free mass: influence of body weight. Am J Clin Nutr 1999;70:1025e Houston DK, Stevens J, Cai J. Abdominal fat distribution and functional limitations and disability in a biracial cohort: the Atherosclerosis Risk in Communities Study. Int J Obes (Lond) 2005;29:1457e Balkau B, Deanfield JE, Despres JP, Bassand JP, Fox KA, Smith SC, et al. International Day for the Evaluation of Abdominal Obesity (IDEA): a study of waist circumference, cardiovascular disease, and diabetes mellitus in 168,000 primary care patients in 63 countries. Circulation 2007;116:1942e Pischon T, Boeing H, Hoffmann K, Bergmann M, Schulze MB, Overvad K, et al. General and abdominal adiposity and risk of death in Europe. N Engl J Med 2008;359:2105e Hughes VA, Roubenoff R, Wood M, Frontera WR, Evans WJ, Fiatarone Singh MA. Anthropometric assessment of 10-y changes in body composition in the elderly. Am J Clin Nutr 2004;80:475e Szulc P, Claustrat B, Marchand F, Delmas PD. Increased risk of falls and increased bone resorption in elderly men with partial androgen deficiency: the MINOS study. J Clin Endocrinol Metab 2003;88:5240e7. Please cite this article in press as: Genton L, et al., Body composition changes over 9 years in healthy elderly subjects and impact of physical activity, Clinical Nutrition (2011), doi: /j.clnu

117 APPENDIX 9

118 OBES SURG (2010) 20: DOI /s y CASE REPORT Parenteral Nutrition Independence in a Patient Left with 25 cm of Ileum and Jejunum: A Case Report Laurence Genton & Patrizia Nardo & Olivier Huber & Claude Pichard Received: 15 December 2009 / Accepted: 28 January 2010 / Published online: 26 February 2010 # Springer Science+Business Media, LLC 2010 Abstract A 44-year-old woman with a history of Roux-en-Y gastric bypass (RYGBP) suffered small bowel volvulus. She was left post-operatively with an intact duodenum, 25 cm of jejunum and ileum, and a colon in continuity, a situation synonymous to short bowel syndrome. This report describes her surgical, medical and nutritional follow-up until complete weaning of parenteral nutrition despite of her very short remnant small bowel and persistently low citrullinemia. The discussion aims at demonstrating the rarity of these complications after RYGBP according to the literature. Furthermore, it challenges the validity of the present markers of parenteral nutrition independence (remnant small bowel length, citrullinemia) in case of short bowel syndrome. Keywords Short bowel syndrome. Gastric bypass. Parenteral nutrition. Body composition. Citrulline Introduction L. Genton : P. Nardo : C. Pichard (*) Clinical Nutrition, Geneva University Hospital, Rue Gabrille-Perret-Gentil 4, 1211 Geneva 14, Switzerland claude.pichard@unige.ch O. Huber Visceral and Transplantation Surgery, Geneva University Hospital, Rue Gabrille-Perret-Gentil 4, 1211 Geneva 14, Switzerland Short bowel syndrome (SBS) is a rare condition with an incidence of two per million and a prevalence of two to four per million subjects in most western countries [1]. It generally results from surgical bowel resection related to recurrent Crohn s disease, mesenteric arterial infarct, venous thrombosis, volvulus, trauma or tumour. A functional SBS may also occur in case of severe malabsorption with intact bowel length as in radiation enteritis, pseudo-obstruction, refractory sprue or congenital villous atrophy [1]. Patients present with permanent or transient intestinal failure leading to water, mineral and macronutrient malabsorption [2] and thus often require long-term parenteral nutrition (PN). Messing et al. showed that the cut-off values for permanent intestinal failure, defined as PN dependency over 2 years duration, was 100 cm in case of end-jejunostomy and 70 cm in case of jejunocolic or jejuno-ileocolic anastomosis [3]. This article reviews the case of a woman who reached PN independence with an intact duodenum, about 25 cm jejunum and ileum and a colon in continuity. Case Report A 44-year-old women with a BMI of 45 kg/m 2 benefited from bariatric surgery in July Her medical history was otherwise not relevant. She underwent a laparoscopic proximal Roux-en-Y gastric bypass (RYGBP) and a cholecystectomy for pigment stone (day 561) (Fig. 1a). Biopsies showed a discrete macrovacuolar steatosis of the liver, a chronic cholecystitis and cholesterolosis. On August 2007, she complained of abdominal post-prandial cramps and underwent a laparoscopic section of an adhesion in regard of the jejuno-jejunal anastomosis. Massive Enterectomy, End-jejunostomy and Nutritional Support (day 0) In February 2008, while she was in Bolivia, she suffered from a hypovolemic shock related to an extended small bowel necrosis following a Roux limb volvulus. She

119 OBES SURG (2010) 20: Fig. 1 Abdominal status after gastric bypass (a), at admission to the Geneva University Hospital after intestinal volvulus in Bolivia (b), after the laparoscopy performed at the Geneva University Hospital (c), after establishment of jejuno-ileo-sigmoid continuity (d) and finally after establishment of ileo-colic continuity (e) underwent subtotal enterectomy with placement of a gastrostomy and a jejunostomy (day 0) and was transferred to the University Hospital of Geneva 9 days later. At day 18, she underwent an exploratory laparotomy, which showed an intact colon, the presence of the appendix and 5 cm of distal ileum closed proximally. The remaining jejunum and proximal ileum measured about 25 cm, were anchored distally to the skin as jejunostomy and proximally to the duodenum and distal gastric pouch, which was equipped with a feeding gastrostomy (Fig. 1b). The proximal gastric pouch had been left with one cm of small bowel. The one cm of jejunum attached to the proximal gastric pouch was removed and a gastro-gastric anastomosis performed (Fig. 1c). The patient was fed through a central venous catheter since day 10. She weighed 64 kg for a height of 160 cm. PN provided 2,215 kcal daily (1,875 ml, 270 g carbohydrates, 108 g amino acids, 75 g fat; Nutriflex, B. Braun Medical, Crissier, Switzerland). Trace elements (10 ml Addamel, Fresenius Kabi, Stans, Switzerland), vitamins (5 ml Cernevit, Baxter, Volketswil, Switzerland) and glutamine (80 kcal, 134 g L-glutamine, 82 g L-alanine; Dipeptiven, Fresenius Kabi, Stans, Switzerland) were added into the PN bag. On day 31, an indirect calorimetry showed an energy expenditure under PN of 1,630 kcal, estimated at 1,900 kcal with physical mobilisation (Table 1). Thus, PN was decreased to 1,890 kcal daily. Intravenous hydration and electrolytes were supplemented daily according to the losses through the jejunostomy, which varied between 1.5 and 4 l/day, renal function and osmolality. To promote bowel adaptation, the patient received daily 750 kcal of a polymeric fibre-enriched enteral nutrition (500 ml Novasource Energy, Nestlé, Vevey, Switzerland) through the gastrostomy until she resumed oral feeding and gastrostomy could be removed (day 34). To limit jejunostomy output, she received soluble fibre supplementation and esomeprazole. Regarding oral feeding, she was advised to limit intakes of raw fruits and vegetables, add 8 10 g/day salt to her diet and drink high-salt containing water to increase water absorption, dissociate drinking from eating to reduce transit time, limit lactose-containing food as it may induce osmotic diarrhoea and to have at least six small intakes per day. Nutritional support is detailed on Fig. 2. The patient was discharged on day 44 with 1,890 kcal/day nocturnal PN, oral feeding, esomeprazole, buprenorphine for parietal, pain and anticoagulation because of a thrombosis of the mesenteric vein and a right segmental lung emboly discovered fortuitously on a CT-scan. Between day 45 and day 121, the patient remained under the same feeding regimen but received also 0.5 to 2 L NaCl 0.9% per day intravenously according to renal function and osmolality. Evolution of weight and body composition is shown on Fig. 3. Nutritional plasma parameters are presented on Table 2. During the whole follow-up, liver tests and triglycerides remained within normal limits while plasma citrulline remained below limits (between 7 and 10 umol/l; normal values: umol/l). Levels of plasma vitamins were measured every 2 months and supplemented when necessary. Jejuno-ileo-sigmoid Continuity and Nutritional Support (day 122) The patient was again hospitalised on day 122 to establish jejuno-ileal continuity. As the stomy output frequently

120 668 OBES SURG (2010) 20: Table 1 Energy expenditure measured by indirect calorimetry from day 627 (before the Roux-en-Y gastric bypass) until day +405 (weaning of home parenteral nutrition to twice per week) Day Energy expenditure a, kcal VO2, ml/min VCO2, ml/min Respiratory quotient Feeding Quantity of PN, kcal Duration of PN 611 1, Oral feeding a 26 1, Exclusive PN 2,215 Over 24 h 110 1, PN and oral feeding 1,960 From 20 pm to 8 am 251 1, PN and oral feeding 2,295 From 20 pm to 8 am 317 1, PN and oral feeding 2,295 From 20 pm to 8 am 353 1, PN and oral feeding 2,295 From 20 pm to 8 am 405 1, Oral feeding a a When the patient was fed exclusively orally, energy expenditure was measured in fasting conditions exceeded 1,200 ml/day, it was decided to perform a temporary sigmoidostomy and not to establish colonic continuity at once to allow colonic adaptation and avoid anal irritation by frequent stools (Fig. 1d). She was treated from day 111 to 133 with 12 IU (4 mg) human recombinant growth hormone (Saizen, Serono, Geneva, Switzerland) daily to stimulate bowel trophicity. Nocturnal PN was increased in parallel to 2,215 kcal to account for the increased needs due to growth hormone-induced anabolic effect. She was discharged on day 136 and frequency of PN administration could progressively be decreased to 2,215 kcal 5 days/week because of stable weight. The patient was asked to consume an oral powder supplement containing 10 g glutamine (Glutamine Plus, Fresenius, 10 g glutamine and 80 kcal/pack) thrice a day during the PN-free days. Sigmoido-rectal Continuity and Nutritional Support (day 272) She finally underwent reestablishment of continuity between the sigmoid and rectum on day 272 (Fig. 1e). To stimulate bowel trophicity, she was treated again with growth hormone daily (12 UI/day) during 14 days pre- and 5 days post-operatively and frequency of PN was increased from 5 to 7 days/week. The post-operative period was uncomplicated and the patient rapidly discharged. On day 286, PN was decreased to five times per week. Oral glutamine was stopped on day 303. At this time, bowel transit was studied with ingestion of liquid barium and with ingestion of a breakfast containing metallic markers. After barium ingestion, the stomach was emptied within 30 min and colon and rectum largely opacified 2 h later. In contrast, the ingestion of the breakfast led to a surprisingly long intragastric phase as the metallic markers were still in the stomach4hlaterandtheileo-caecal valve was reached only6hlater,suggestingaslowertransittimewithsolid mealscomparedwithliquids. The follow-up was significant for an infection of the implantable chamber (day 326) with Klebsiella pneumonia and Staphylococcus epidermidis. The chamber was excised as its sterilisation failed with 2 weeks of antibiotic treatment (intravenous ceftobiprole and vancomycine-lock), and a new chamber was implanted on day 345. PN was kcal/d days Parenteral nutrition Glutamine iv Enteral nutrition Fig. 2 Nutritional support by enteral and parenteral nutrition since the occurrence of the intestinal volvulus