Alexandra M. Johnstone a, Peter Faber b, Eileen R. Gibney c, Gerald E. Lobley a, R. James Stubbs a, Mario Siervo d, ORIGINAL ARTICLE

Transcription

1 Obesity Research & Clinical Practice (2014) 8, e46 e54 ORIGINAL ARTICLE Measurement of body composition changes during weight loss in obese men using multifrequency bioelectrical impedance analysis and multi-compartment models Alexandra M. Johnstone a, Peter Faber b, Eileen R. Gibney c, Gerald E. Lobley a, R. James Stubbs a, Mario Siervo d, a University of Aberdeen Rowett Institute of Nutrition and Health, Greenburn Road, Bucksburn, Aberdeen AB21 9SB, UK b Department of Anaesthetics, Aberdeen Royal Infirmary, Foresterhill, Aberdeen AB25 2ZD, UK c College of Life Sciences, University College, Dublin, Ireland d Human Nutrition Research Centre, Institute for Ageing and Health, Newcastle University, Campus for Ageing and Vitality, Newcastle on Tyne NE4 5PL, UK Received 21 March 2012; accepted 28 July 2012 KEYWORDS Obesity; Weight loss; Multi-compartment models; Dilution methods; Multi-frequency bioimpedence Summary Background: The accurate measurement of changes in body composition is important to assess the contribution of fat and fat free mass to total body mass change as a measure of the effectiveness of weight loss programmes. Bioelectrical impedance spectroscopy (BIS) is a rapid and non-invasive technique which could be applied to assess body composition changes. The aim of the study was to evaluate the accuracy of the BIS for the measurement of fat mass (FM), total body water (TBW) and extracellular water (ECW) changes induced by different degrees of caloric deficit in obese men. Methods: Three groups of six, obese men participated in either (i) a total fast (for 6 days); (ii) a VLCD (2.5 MJ/day for 3 weeks); or (iii) LCD (5.2 MJ/day for 6 weeks). FM was measured using a 4-compartment (4-C) model. TBW and ECW were determined by dilution methods, respectively. TBW, ECW and FM were also assessed with BIS. Results: Body weight loss in the fasting group was 6.0 ± 1.3 kg over 6 days; the VLCD group lost 9.2 ± 1.2 kg over 21 days and the LCD group lost 12.6 ± 2.4 kg over 42 days. BIS underestimated FM changes (bias = 3.3 ± 3.8 kg) and overestimated changes in TBW and ECW by +1.8 ± 4.8 kg and +2.3 ± 6.4 kg, respectively. The measurement error was consistently larger in the fasting group and the magnitude of the bias interacted significantly with the rate of weight loss. Corresponding author. Tel.: address: mario.siervo@ncl.ac.uk (M. Siervo) X/$ see front matter 2012 Asian Oceanian Association for the Study of Obesity. Published by Elsevier Ltd. All rights reserved.

2 Bioelectrical impedance spectroscopy and body composition changes e47 Conclusion: Rapid weight loss affects the accuracy of the BIS in detecting changes in body composition. A careful interpretation of the results is needed when sizable changes in body water compartments occurs Asian Oceanian Association for the Study of Obesity. Published by Elsevier Ltd. All rights reserved. Introduction The assessment of changes in body composition such as fat mass (FM) or total body water (TBW) and their contribution to the total amount of weight lost are important to provide a less confounded measure of the effectiveness of weight loss (WL) interventions [1]. Multi-compartment models and isotope dilution methods are considered the reference methods for assessment of FM, TBW and extracellular water (ECW), respectively [2]. The main advantage of multi-compartment models is that they integrate information obtained from single measurements (body density, total mineral mass, total body water) to reduce the number of assumptions made on the stability of body compartments (hydration, body mineral density) [3]. These models are characterised by a limited application in clinical practice as they do not provide immediate result outputs, are expensive and require advanced analytical and technical expertise [2]. Bioelectrical impedance analysis (BIA) has instead all the prerequisites (safe, inexpensive, non-invasive, rapid, portable) for being potentially used in clinical settings to measure individual changes in body composition [4]. A recent review, however, has questioned the accuracy and precision of this technique for the assessment of body compartments at the individual level [5]. For example, the standard error of the estimate (SEE) for regression of TBW assessed using BIA against isotope dilution in healthy individuals can vary from 1.3 to 3.8 kg [5]. In addition, the accuracy of BIA may be affected by changes in body geometry, hydration, age or ethnicity [6]. The application of modelling techniques and mixture theories (Cole Cole method and Hanai theory) to multi-frequency bioelectrical impedance data (bioelectrical impedance spectroscopy (BIS)), has improved the estimation of body fluid compartments, maintaining the user-friendly features of the BIA [7]. The improved accuracy of the BIS has been reported in samples with different clinical and nutritional characteristics [8 12]. Here we compared the accuracy of bioelectrical spectroscopy (BIS) to body composition reference methods for the measurement of body composition changes in obese subjects following three different weight loss programmes (fasting, very low calorie diet (VLCD) and low calorie diet (LCD)). Methods Subject characteristics The data reported in this manuscript are derived from a secondary analysis of an experimental study conducted between 1999 and 2001 to investigate the effects of different levels of negative energy balance on the physiology of body weight regulation. Elements of the characteristics of some of the subject groups and their responses to the dietary interventions have been presented previously [13 15] but are repeated here to provide an overall picture. Eighteen (6 subjects in each WL group), healthy, obese (BMI range: kg/m 2 ) male subjects, aged between 20 and 55 years, were recruited. Subjects were included if they were not on any special or prescribed diet; were non-smokers; had stable weight (weight change of no more than 2 kg in the previous 3 months); were healthy based on medical examination, blood tests and electrocardiogram. The volunteers took no regular prescribed medication, vitamin or mineral supplements. This study was approved by the by the Grampian Research Ethics Committee. Written informed consent was obtained from all subjects. Study design In the first group, subjects were fasted for 6 days to lose a nominal 5% original body weight (OBW), this magnitude of WL was set by ethical constraints. The second group consumed a very-low-calorie diet (VLCD) for 3 weeks to lose a nominal 10% OBW and

3 e48 A.M. Johnstone et al. the third group followed a low-calorie-diet regime (LCD) for 6 weeks, also to lose a nominal 10% OBW. Study protocols Each study had a similar design. On days 1 6 (termed baseline period), subjects consumed a fixed maintenance diet (13% protein, 30% fat and 57% CHO), calculated to meet energy requirements (estimated at 1.6 measured resting metabolic rate, RMR). Then followed the period of WL. Throughout the study, participants were residential in the Human Nutrition Unit at the Rowett Research Institute, Aberdeen (UK). Energy intake Energy intake (EI) was measured daily, based on the recorded weighed intakes of food and drinks. During fasting, the participants had access to water only. The VLCD comprised: daily weight 642 g, energy 2.55 kj/g, protein 49.4 g (0.84 kj/g), carbohydrate 52.8 g (0.85 kj/g), and fat 23.1 g (0.85 kj/g). The LCD comprised: daily weight 1260 g, energy 5.0 kj/g, protein 50.3 g (0.89 kj/g), carbohydrate g (2.61 kj/g), and fat 45.4 g (0.85 kj/g). Measurement of body weight and body composition changes Anthropometry Subject height was measured once to the nearest 0.1 cm. Each morning body weight was measured to the nearest 50 g using a digital scale before eating and after voiding, with subjects wearing light clothes. Fat mass The four-compartment model of body composition [16] was implemented to assess longitudinal changes in FM according to the following formula: FM (kg) = BV TBW A Wt, where BV is body volume (l), TBW is total body water (l), A is ash (kg) and Wt is body weight (kg). Total body water (TBW-kg) was measured by deuterium dilution (D 2 O). Bone mineral mass (A-kg) was measured by dual energy X-ray absorptiometry (DXA). Body volume and density was measured using a system of air displacement plethysmography. Detailed descriptions of the methods have been provided elsewhere [17]. Total body water and extracellular water Total body water (TBW-kg) was measured by deuterium dilution (D 2 O) as described by Pullicino et al. [18]. Plasma samples were prepared for enrichment analysis by the zinc reduction method [19], and analysed using a VG SIRA-10 (VG Isogas, Cheshire, England) gas isotope ratio mass spectrometer. TBW was corrected for the fraction (4%) of hydrogen that is assumed to exchange with non-aqueous hydrogen. The addition of sodium bromide (NaBr) to the mixture allows extracellular water (ECW) to be measured since the bromide mixes only with the extracellular compartments. The pre and post dose concentrations of NaBr were determined by the spectrophotometric fluorescein method [20]. The CV of this technique was determined as % by Jennings and Elia [20]. ECW was then calculated using the formulae described by Cornish et al. [21]. ECW (L) = ( ) Br-dose (mmol) 0.90 [Br] plasma (mmol/l) where is the fraction of water in plasma [20], 0.90 is the fraction of the retained bromide dose that is assumed to remain extracellular [22] and 0.95 is the equilibration factor for bromide [23]. Bioelectrical impedance spectroscopy (BIS) TBW and FM were assessed using tetra polar multiple-frequency bioelectrical impedance spectroscopy (SFB3, SEAC, Australia). Subjects were asked to void their bladders and remove any metallic objects before the measurements. Briefly, measurements were taken within 5 10 min of subjects lying supine to reduce body water redistribution. Electrodes were attached to the upper and lower extremities on specific anatomical landmarks. The measurements were performed by the same operator to decrease the measurement error related to the placement of the electrodes. The application of the tetrapolar BIA technique in repeated measurements is supported by the high short-term (1 h, CV = 0.5%) and long-term precision (10 weeks, CV = 2.5%) [24]. The SBF3 instrument obtains measurements on 496 frequencies in the range khz. The instrument provides raw data for each frequency (R and Xc and Z magnitude and phase angle at each frequency, including 50 khz) and the software provided by the manufacturer estimates TBW, ECW, and ICW according to the Cole Cole method and Hanai model (equations are described elsewhere [25,26]). Fat free mass (kg) was calculated

4 Bioelectrical impedance spectroscopy and body composition changes e49 as TBW/0.73 and FM calculated as the difference between body weight and FFM. Statistical analysis Data are shown as mean ± SD. For each WL study, the body composition values were analysed by ANOVA. The fixed effects consisted of method (M), time (T) and their interaction (I). Where either main or interaction effects were significant (p < 0.05) then a post hoc t-test was applied to identify differences between means. A similar statistical approach was applied to examine differences in body composition as measured by method and with values obtained during the baseline period subtracted for each phase ( ). The assumptions of normality and constant variance were always met, based on visual inspection of the residual plots. Linear regression, correlation analysis and the Bland Altman method were used to assess the agreement between the two methods for the measurement of changes in TBW, ECW and FM. The analysis was conducted on the pooled measurements (5% WL and 10% WL phases) performed during the study (n = 30) and divided by weight loss groups (fasting, n = 6: VLCD, n = 12; LCD, n = 12). A paired t-test was used to detect between-method statistically differences. Statistically significant difference was defined as a p value < Data analyses were performed using Genstat 11th (VSN International Ltd, United Kingdom) and SPSS (SPSS 16.0; SPSS, Chicago, IL). Results Body weight (BW) On average, the percentage WL at the intended 5% level were 5.6, 4.9 and 6.8% OBW on the fasting, VLCD and LCD group, respectively. The weight loss achieved at the end of the second weight loss phase was still higher for the LCD group (11.9%) compared to the VLCD group (8.6%). Each intervention induced significant body mass changes relative to baseline (p < 0.001) (Table 1). Fat mass (FM) The changes in FM during fasting, VLCD and LCD assessed by the 4-C model were all significant (p = 0.001). In the fasting group, FM decreased by 3.2 ± 0.5 kg at the end of the 5% WL phase. The VLCD and LCD groups lost 7.4 ± 0.8 kg and 11.5 ± 2.0 kg at the end of 10% WL, respectively (Table 1). Figure 1 Bland Altman analysis to test the agreement between changes in body composition (FM, TBW, ECW) relative to baseline measured by 4 compartment model and dilution methods (reference methods) with multi-frequency bioelectrical spectroscopy (BIS). (a) Correlation between the changes in body composition for each variable measured by the reference method and the BIS. (b) Visual representation of the bias between the BIS and reference for each variable investigated. Solid lines represent the average bias for each body composition measurement. The Bland Altman method is described in section Methods and the width of the limits of agreements (±2SD) for each measurement is shown in Table 2. The increase in FM measured by BIS in the fasting group at the end of 5% WL ( = +1.8 ± 2.3 kg) was unexpected (Table 1). The performance of the BIS method improved in the LCD group as the differences were smaller compared to the other two weight loss interventions. However, FM was still significantly underestimated compared to the 4-C model when the FM measurements (absolute and changes) were pooled together (Tables 1 and 2). The Bland Altman analysis (Table 2 and Fig. 1b) confirmed the poor accuracy of the BIS in measuring FM as BIS was characterised by a large measurement bias ( FM = 3.3 kg) and wide limits of agreement (±2SD = 3.8 kg). BIS consistently underestimated FM changes associated with weight loss and the bias was directly associated with the degree of negative

5 e50 A.M. Johnstone et al. Table 1 Body composition changes in obese subjects during weight loss in the total sample and divided by weight loss programme (fasting, very low calorie diet (VLCD) and low calorie diet (LCD)). Fat mass (FM), total body water (TBW) and extracellular water (ECW), measured by a four-compartment model (4-C model), deuterium dilution (D 2 O) and sodium bromide dilution (NaBr) respectively, were compared to multi-frequency bioelectrical spectroscopy (BIS) at the end of each phase. Baseline 5% WL 10% WL Time (T) Method (M) Fasting (n = 6) Weight (kg) (11.5) 6.0 (1.3) p < FM 4-C (kg) 38.7 (6.5) 3.2 (0.5) A: p = A: p = A: p = FM BIS (kg) 26.9 (5.9) +1.8 (2.3) : p = : p < : p < TBW D2 O (kg) 50.7 (4.6) 3.2 (1.2) A: p = A: p = A: p = TBW BIS (kg) 58.6 (5.4) 4.5 (4.8) : p = : p = : p = ECW NaBr (kg) 20.9 (2.4) +1.5 (3.1) A: p = A: p = A: p = 0.03 ECW BIS (kg) 25.3 (2.8) 4.1 (0.7) : p = : p < : p < VLCD (n = 6) Weight (kg) (14.9) 5.2 (0.8) 9.2 (1.2) p < FM 4-C (kg) 45.3 (9.9) 3.6 (1.0) 7.4 (0.8) A: p = A: p < A: p = FM BIS (kg) 31.0 (7.2) 0.8 (1.0) 5.5 (0.7) : p < : p < : p < TBW D2 O (kg) 46.6 (5.1) 1.2 (0.6) 1.8 (1.4) A: p = A: p < A: p = TBW BIS (kg) 55.5 (7.3) 3.3 (1.8) 2.3 (1.2) : p < : p = : p = ECW NaBr (kg) 22.4 (3.1) 2.0 (1.5) 1.8 (2.8) A: p = A: p = A: p = ECW BIS (kg) 25.1 (3.6) 2.5 (1.6) 2.0 (0.6) : p = : p = : p = LCD (n = 6) Weight (kg) (10.0) 7.2 (1.5) 12.6 (2.4) p < FM 4-C (kg) 40.7 (8.7) 5.9 (1.4) 11.5 (2.0) A: p = 0.03 A: p < A: p = FM BIS (kg) 26.7 (5.5) 2.8 (1.1) 7.4 (3.2) : p < : p < : p = TBW D2 O (kg) 48.2 (3.1) 0.3 (0.9) 0.6 (0.2) A: p = A: p < A: p = TBW BIS (kg) 57.6 (4.6) 3.2 (0.9) 3.4 (0.5) : p < : p < : p < ECW NaBr (kg) 20.1 (1.0) 0.2 (1.6) +2.4 (2.6) A: p = A: p = A: p = ECW BIS (kg) 23.9 (2.4) 1.9 (0.4) 1.5 (0.5) : p = : p < : p = ALL (n = 18) a Weight (kg) (11.6) 6.1 (1.4) 10.9 (2.5) p < FM 4-C (kg) 41.6 (8.5) 4.2 (1.5) 9.4 (2.6) A: p = A: p < A: p = FM BIS (kg) 28.2 (6.2) 0.5 (2.5) 6.4 (2.3) : p < : p < : p < TBW D2 O (kg) 48.5 (4.5) 1.5 (1.5) 1.2 (1.2) A: p = 0.05 A: p < A: p = TBW BIS (kg) 57.2 (5.7) 3.6 (2.9) 2.8 (1.0) : p < : p < : p = ECW NaBr (kg) 21.1 (2.4) 0.05 (2.6) 0.2 (3.4) A: p = A: p = A: p = ECW BIS (kg) 24.8 (2.8) 2.8 (1.4) 1.7 (0.6) : p = : p < : p = Data are shown as mean (SD). A = analysis carried out on absolute values; = absolute changes relative to baseline and also indicates statistical analysis carried on changes in TBW and FM relative to baseline. For each weight loss (WL) study, the body composition values were analysed by ANOVA. The fixed effects consisted of method (M), time (T) and their interaction (T M). Significant results are highlighted in bold. a Measurements were available from twelve subjects at the end of the 10% WL phase. T M energy balance. The correlation of BIS with the 4-C model was good for the VLCD and LCD groups but very poor and not significant for the fasting group (Table 2 and Fig. 1a). The measurement bias was non-differential as it was not significantly associated with changes in FM (r = 0.24, p = 0.06) (Fig. 1b). Changes in TBW were significant in the three weight loss interventions whereas changes in ECW were significant only in the VLCD group (Table 1). The BIS method consistently overestimated TBW and the error was larger in the LCD group (Table 1). Similarly, the BIS performed poorly compared to the reference method (sodium bromide dilution) in the assessment of ECW as the magnitude (T) and the direction (T M) of the ECW changes were significantly different in the fasting and LCD groups (Table 1). The agreement between the reference method and BIS in measuring TBW and ECW changes is shown in Table 2 and Fig. 1b. The average bias for the determination of changes in TBW was larger in the LCD group (+2.7 kg) but the limits

6 Bioelectrical impedance spectroscopy and body composition changes e51 Table 2 Bland Altman analysis to test the agreement between changes ( ) in fat mass (FM), extracellular water (ECW) and total body water (TBW), measured respectively by 4 compartment model (4-C), sodium bromide dilution and deuterium dilution, with multi-frequency bioelectrical spectroscopy (BIS). 4-Compartment model Bromide dilution Deuterium dilution FM (kg) ECW (kg) TBW (kg) Bias ±2SD SEE r Bias ±2SD SEE r Bias ±2SD SEE r All 3.3 c c c Fasting 5.1 b a VLCD 2.3 c LCD 3.4 c b c Bias = BIS-reference method; plus sign (+) = overestimation of changes; minus sign ( ) = underestimation of changes; SD = standard deviation; SEE = standard error of estimate; r = coefficient of correlation. The analysis was conducted on all the body composition measurements (5% WL and 10% WL phases) performed during the study (All, n = 30) and divided by weight loss groups (fasting, n = 6; VLCD, n = 12; LCD, n = 12). Statistical difference between BIS and reference methods were tested by paired t test. a p < b p < c p < of agreement (±2SD) were larger in the fasting group (2SD = ±8.6 kg). The correlation between BIS and deuterium dilution was poor (r = 0.23) for the pooled data (Fig. 1a). Changes in TBW were on average overestimated by the BIS and a significant differential bias was found as the measurement bias was directly associated with TBW changes (r = 0.49, p < 0.001) (Fig. 1b). Overall BIS overestimated changes in ECW by +2.3 kg but limits of agreements were wide (2SD = ±6.4 kg) and correlation very poor (r = 0.09) (Fig. 1a). The error became larger during fasting (+5.7 kg) with very wide limits of agreement (2SD = ±7.2 kg) (Table 2). Fig. 1b shows how changes in ECW were overestimated by the BIS (Fig. 1b) and a significant differential bias was found (r = 0.66, p < 0.001). an unexpected increase in FM by BIS was observed in the fasting group at the end of 5% WL. The BIS lacked both accuracy and precision. Specifically, the overestimation of changes in TBW and the underestimation of changes in FM persisted during the study (accuracy) whereas the magnitude of the bias between BIS and reference methods was changing during the study as we observed an amplification of the bias towards the end of each weight loss intervention (precision). Cox-Rejiven et al. [27] compared the performance of the BIS method (where a different device was used) to the isotope dilution method to measure TBW. The characteristics of their sample were slightly different as BMI ranged from 23 to 52 kg/m 2 Impedance The extent of the differences in impedance measurements by BIS in each weight loss group is shown in Fig. 2. The only significant change in impedance values at 50 khz occurred in the fasting group (p = 0.02), which is probably a reflection of the important fluid shifts associated with the rapid weight loss. Discussion The error associated with the measurement of body water (TBW, ECW) and FM changes by BIS was directly associated with the rate of weight loss. Measurement error and limits of agreement were consistently larger in the fasting group and Figure 2 Impedance (50 khz) mean values measured by BIS at the end of each phase for each weight loss group. Error bars are standard errors. Bioimpedance values only increased significantly after 5% weight loss (5% WL) in the fasting group.

7 e52 A.M. Johnstone et al. and males and females were included, which may have increased the variability of the measurements. Their SEE for the measurement of TBW compared with isotope dilution (deuterium) was 2.2 kg, close to that observed in our study (pooled SEE for TBW = 2.4 kg). In contrast, the SEE for TBW in the fasting group was considerably higher (4.6 kg) than in the study of Cox-Rejiven et al. [27]. Furthermore, the SEE values were reduced in the VLCD and LCD groups, suggesting that the rapid and acute fluid shifts affected the estimation of TBW by BIS. Four studies have explored the validity of BIS in obese subjects and three investigated changes in body composition during weight loss [8,27 29]. Two studies followed obese subjects undergoing bariatric surgery to assess TBW changes by BIS and deuterium dilution. Both studies concluded that the error associated with BIS was worse with increasing loss of FM and the individual variability in the measurements would limit the application of BIS in the clinical setting for extremely obese patients [27,28]. In contrast, a recent study compared the BIS to the DXA for the measurement of FM changes provided more positive results in 24 overweight and obese females (BMI = 36.4 ± 4.3 kg/m 2 ) following a 10-week weight loss programme ( 30% caloric deficit). The authors concluded that BIS accurately assessed changes in FM and BIS provided better estimates of body composition than single-frequency BIA [29] and this suggest that BIS may be more accurate in interventions characterised by a slower weight loss. Nonetheless, although group means were similar across the various methods, the associated variance at the individual level was still high and therefore we believe that the BIS approach would not be appropriate to measure body composition at individual level. In fact the standard deviation of FM loss, measured by DXA, was ±2.1 kg for an average weight loss of 3.2 kg, which means that a certain number of subjects (not provided in the paper) could have a FM change lower than 1 kg, which would certainly be close to the sensitivity threshold of the DXA method for FM changes and go certainly beyond the detection limit for FM changes of the BIS method [30]. In our study, the error was larger during the fasting period. This result may have been expected considering the rapid and significant fluid shifts seen both during rapid weight loss. This effect seems to be confirmed by the significant increase in impedance values at 50 khz which are possibly related to the loss of conductive elements like water and electrolytes during rapid weight loss. This may have led to the unexpected measurement of an increase in FM by BIS after 5% WL. The BIS performed slightly better in the VLCD and LCD interventions, with the direction of the changes compatible with the 4-C model estimates. Some technical and theoretical assumptions included in the software provided by the manufacturer for the analysis of the multi-frequency data may have influenced the results. The software integrates a modelling approach (Cole Cole method) with emulsion theories (Hanai model) to determine ECW, ICW and TBW. Several authors have challenged the applicability of the resistivity coefficients used in the Hanai model for the derivation of ECW and ICW as they were in a sample of normal weight subjects [8,31]. Previous attempts have tried to correct for this intrinsic error of the equations by deriving new coefficients for obese subjects but the authors still concluded that, despite an increased accuracy of the BIS, the error was still large and they still did not justify using the BIS at individual level [8]. We have not attempted in this analysis to derive any new resistivity coefficient to correct the equations for body size, body geometry or changes in hydration status as we wanted to provide information on the reliability of the BIS as it would be routinely used in weight loss clinics. The study has some limitations which are here discussed. The manuscript reports results from a secondary analysis of data having as primary objective the investigation of the effects of different levels of negative energy balance of body composition and energy expenditure. Nonetheless, the study offered the opportunity to test the performance of BIS in three different weight loss interventions characterised by different rates of tissue mobilisation. This could provide relevant information on the application of BIS in clinical practice in particular when more intensive weight loss programmes are used in obese patients (meal replacement, very low calorie diet, intermittent fasting). The secondary nature of the analysis is also reflected in the small sample size of each weight loss group for the evaluation of BIS performance. However, results were not substantially different when study where pooled together. In addition our results confirm previous results on the modest accuracy of the BIS at individual level for the determination of body composition changes during weight loss. The results are discussed in the context of previous weight loss studies employing the BIS method and a discussion of the accuracy of single-frequency BIA has been purposively omitted. However, the accuracy of the single-frequency BIA during weight loss has been previously reviewed with very similar conclusions, i.e., the validity of the technique is limited for the assessment of body composition changes at individual level [5].

8 Bioelectrical impedance spectroscopy and body composition changes e53 In conclusion, the study showed that performance of the BIS was influenced by the rate of weight loss which had important effects on the size of the measurement error. Furthermore, the large limits of agreement seen in the fasting group warrant careful consideration of sample size when this technique is used to assess body composition changes during weight loss interventions including fasting, VLCD or bariatric surgery where rapid body mass and fluid losses are expected to predominate. Author s contributions MS analysed the data and wrote the manuscript; PF, ERG, AMJ designed the study and collected the data; GEL and RJS designed study; all authors read and approved the findings of the study. Conflict of interest None declared. Acknowledgements This work was supported by funding from Scottish Executive Environment and Rural Affairs Department (SEERAD) and a grant from Slimming World, Alfreton, UK. Statistical advice was provided by Dr Grietje Holtrop, Biomathematics and Statistics Scotland (BioSS). References [1] Prentice AM, Goldberg GR, Jebb SA, Black AE, Murgatroyd PR, Diaz EO. Physiological responses to slimming. Proc Nutr Soc 1991;50(2): [2] Siervo M, Jebb SA. Body composition assessment: theory into practice: introduction of multicompartment models. IEEE Eng Med Biol Mag 2010;29(1): [3] Wang Z, Pierson Jr R, Heymsfield S. The five-level model: a new approach to organizing body-composition research. Am J Clin Nutr 1992;56(1): [4] Kyle UG, Bosaeus I, De Lorenzo AD, Deurenberg P, Elia M, Gómez JM, et al. Bioelectrical impedance analysis part I: review of principles and methods. Clin Nutr 2004;23(5): [5] Kyle UG, Bosaeus I, De Lorenzo AD, Deurenberg P, Elia M, Gómez JM, et al. Bioelectrical impedance analysis part II: utilization in clinical practice. Clin Nutr 2004;23(6): [6] Dehghan M, Merchant A. Is bioelectrical impedance accurate for use in large epidemiological studies? Nutr J 2008;7(1):26. [7] Earthman C, Traughber D, Dobratz J, Howell W. Bioimpedance spectroscopy for clinical assessment of fluid distribution and body cell mass. Nutr Clin Pract 2007;22(4): [8] Cox-Reijven P, Soeters PB. Validation of bio-impedance spectroscopy: effects of degree of obesity and ways of calculating volumes from measured resistance values. Int J Obes (Lond) 2000;24: [9] Cox-Reijven PLM, van Kreel B, Soeters PB. Bioelectrical impedance measurements in patients with gastrointestinal disease: validation of the spectrum approach and a comparison of different methods for screening for nutritional depletion. Am J Clin Nutr 2003;78(6): [10] Earthman CP, Matthie JR, Reid PM, Harper IT, Ravussin E, Howell WH. A comparison of bioimpedance methods for detection of body cell mass change in HIV infection. J Appl Physiol 2000;88(3): [11] Lof M, Forsum E. Evaluation of bioimpedance spectroscopy for measurements of body water distribution in healthy women before, during, and after pregnancy. J Appl Physiol 2004;96(3): [12] Fenech M, Maasrani M, Jaffrin MY. Fluid volumes determination by impedance spectroscopy and hematocrit monitoring: application to pediatric hemodialysis. Artif Organs 2001;25(2): [13] Faber P, Johnstone AM, Gibney ER, Elia M, Stubbs RJ, Roger PL, et al. The effect of rate and extent of weight loss on urea salvage in obese male subjects. Br J Nutr 2003;90(1): [14] Johnstone AM, Faber P, Andrew R, Gibney ER, Elia M, Lobley G, et al. Influence of short-term dietary weight loss on cortisol secretion and metabolism in obese men. Eur J Endocrinol 2004;150(2): [15] Siervo M, Faber P, Gibney ER, Lobley GE, Elia M, Stubbs RJ, et al. Use of the cellular model of body composition to describe changes in body water compartments after total fasting, very low calorie diet and low calorie diet in obese men. Int J Obes (Lond) 2010;34(5): [16] Fuller N, Laskey MA, Elia M. Assessment of the composition of major body regions by dual-energy X-ray absorptiometry (DEXA), with special reference to limb muscle mass. Clin Physiol 1992;12: [17] Johnstone AM, Horgan GW, Murison SD, Bremner DM, Lobley GE. Effects of a high-protein ketogenic diet on hunger, appetite, and weight loss in obese men feeding ad libitum. Am J Clin Nutr 2008;87(1): [18] Pullicino E, Coward WA, Stubbs RJ, Elia M. Bedside and field methods for assessing body composition: comparison with the deuterium dilution technique. Eur J Clin Nutr 1990;44(10): [19] Wong WW, Lee LS, Klein PD. Deuterium and oxygen-18 measurements on microliter samples of urine, plasma, saliva, and human milk. Am J Clin Nutr 1987;45(5): [20] Jennings G, Elia M. Automated assay of plasma bromide after a single deproteinization step. Clin Chem 1996;42(8 Pt. 1): [21] Cornish BH, Ward LC, Thomas BJ, Jebb SA, Elia M. Evaluation of multiple frequency bioelectrical impedance and Cole Cole analysis for the assessment of body water volumes in healthy humans. Eur J Clin Nutr 1996;50(3): [22] Bell EF, Ziegler EE, Forbes GB. Corrected bromide space. Pediatr Res 1984;18(4): [23] Wang Z, Heshka S, Wang J, Gallagher D, Deurenberg P, Chen Z, et al. Metabolically active portion of fat-free mass: a cellular body composition level modeling analysis. Am J Physiol Endocrinol Metab 2007;292(1):E49 53.

9 e54 A.M. Johnstone et al. [24] Evans WD, McClagish H, Trudgett C. Factors affecting the In vivo precision of bioelectrical impedance analysis. Appl Radiat Isot 1998;49(5 6): [25] De Lorenzo A, Andreoli A, Matthie J, Withers P. Predicting body cell mass with bioimpedance by using theoretical methods: a technological review. J Appl Physiol 1997;82(5): [26] Matthie J, Zarowitz B, De Lorenzo A, Andreoli A, Katzarski K, Pan G, et al. Analytic assessment of the various bioimpedance methods used to estimate body water. J Appl Physiol 1998;84(5): [27] Cox-Reijven PL, van Kreel B, Soeters PB. Accuracy of bioelectrical impedance spectroscopy in measuring changes in body composition during severe weight loss. JPEN J Parenter Enteral Nutr 2002;26(2): [28] Mager JR, Sibley SD, Beckman TR, Kellogg TA, Earthman CP. Multifrequency bioelectrical impedance analysis and bioimpedance spectroscopy for monitoring fluid and body cell mass changes after gastric bypass surgery. Clin Nutr 2008;27(6): [29] Thomson R, Brinkworth GD, Buckley JD, Noakes M, Clifton PM. Good agreement between bioelectrical impedance and dual-energy X-ray absorptiometry for estimating changes in body composition during weight loss in overweight young women. Clin Nutr 2007;26(6): [30] Ellis KJ. Human body composition: in vivo methods. Physiol Rev 2000;80(2): [31] Moissl UM, Wabel P, Chamney PW, Bosaeus I, Levin NW, Bosy-Westphal A, et al. Body fluid volume determination via body composition spectroscopy in health and disease. Physiol Meas 2006;27(9): Available online at