点击显示 收起
1 From The UCL Institute of Hepatology, Hampstead Campus, Royal Free and University College Medical School, London, United Kingdom (MYM and AMM), and The MRC Dunn Clinical Nutrition Centre, Cambridge, United Kingdom (GJ, ME, and NJF)
2 The British Dietetic Association General and Educational Trust contributed to the cost of the deuterium isotope. 3 Reprints not available. Address correspondence to MY Morgan, The UCL Institute of Hepatology, Hampstead Campus, Royal Free and University College Medical School, University College London, Rowland Hill Street, Hampstead, London NW3 2PF, United Kingdom. E-mail: mymorgan{at}medsch.ucl.ac.uk.
ABSTRACT
Background: Most techniques for measuring body composition are based on 2-component models (2-CMs) and depend on assumptions relating to the constancy of the density (DFFM) and hydration fraction (HFFFM) of fat-free mass (FFM).
Objectives: The objectives were to determine whether these assumptions are systematically violated in patients with cirrhosis and to assess the validity of the estimates of body composition obtained in these patients by using 2-CM techniques.
Design: Body composition was assessed by using a 4-component model (4-CM), which was based on data obtained from densitometry, deuterium dilution, and dual-energy X-ray absorptiometry, in 20 patients with cirrhosis who had no evidence of fluid retention and in 20 pair-matched healthy control subjects. The results were compared with those obtained by using "reference" and "bedside" 2-CM techniques.
Results: The mean (±SD) DFFM was significantly lower in the patients with cirrhosis (1.091 ± 0.008 compared with 1.100 ± 0.006 kg/L; P < 0.001); no significant difference in HFFFM was observed between the patients and control subjects (74.5 ± 2.6 compared with 73.5 ± 2.1), although there was greater variability in the patients. Significant differences were observed in the body-composition variables obtained by using the "reference" 2-CM techniques compared with the 4-CM—the 95% limits of agreement in the patients with cirrhosis exceeded 5% body fat and 3 kg FFM; the corresponding values for the "bedside" 2-CM techniques were 11% body fat and 7.5 kg FFM.
Conclusions: Assumptions relating to the constancy of the DFFM and HFFFM are violated in patients with cirrhosis. Thus, standard 2-CM techniques provide inaccurate body composition estimates in this patient population.
Key Words: Anthropometric measures • bioelectrical impedance analysis • body composition • cirrhosis • 4-component model • densitometry • density of fat-free mass • deuterium dilution • dual-energy X-ray absorptiometry • fat-free mass • healthy control subjects • hydration of fat-free mass • near-infrared interactance • percentage fat mass • urinary creatinine
INTRODUCTION
Patients with cirrhosis are frequently malnourished (1, 2), and this has a detrimental effect on outcome (3-5). The changes in body composition that occur in this patient population are, however, poorly researched and inadequately documented (6-8).
The most frequently used "reference" techniques for assessing body composition are densitometry, deuterium dilution, and dual-energy X-ray absorptiometry (DXA) (9-11). All 3 methods are based, conceptually, on 2-component models (2-CMs), which distinguish the chemically distinct fat mass from the fat-free mass (FFM; essentially water, protein, and mineral). These, and the so-called "bedside" assessment techniques—eg, skinfold thickness anthropometry and bioelectrical impedance analysis (BIA), which are also based on 2-CMs—are critically dependent on the constancy of the relations among the gross components of the FFM and on assumptions relating to its density (DFFM) and hydration fraction (HFFFM). Over- or underestimates of body composition variables will be obtained if these assumptions are invalid.
In recent years, multicomponent techniques have been developed in an attempt to measure FFM constituents more directly and with greater certainty. The 3-CM, which defines the chemically distinct water, fat, and protein plus mineral (fat-free dry matter) compartments, is based on measurements obtained from densitometry and water dilution (12). The 4-CM, which defines the chemically distinct water, fat, protein and mineral, is based, similar to the 3-CM, on measurements of body density (BD) and total body water (TBW) but also incorporates a direct measure of bone mineral content (BMC) obtained with the use of DXA (13, 14). Although the 4-CM depends on an assumed constant ratio of BMC to nonosseous mineral (13, 15), it is fairly robust to major changes in this ratio (14). Use of multicomponent models will obviate the need for some of the assumptions inherent in the use of single techniques, thereby increasing the accuracy of the assessment, without loss of precision (7, 14).
Many of the assumptions made in the assessment of body composition variables in healthy persons are violated in patients with cirrhosis because of the abnormalities in fluid homeostasis and compartmentalization (16), protein metabolism (17), and bone modeling and remineralization (18) that characterize this condition. Nevertheless, few attempts have been made to obtain accurate body composition data in this patient population by using multicomponent models (19-22), even though these data are needed for a number of reasons: first, to allow definition of the pattern of tissue loss; this may provide insights into the disease process and information on outcome and prognosis; it may also affect approaches to nutritional therapy, its monitoring and evaluation; second, to obtain accurate definition of the metabolically active tissue compartments so that variables such as energy expenditure and protein synthesis can be correctly referenced (6); and third, to validate the techniques currently used to assess nutritional status so that their use can be optimized in the clinical setting.
The aim of the present study was to assess body composition in patients with cirrhosis by using a 4-CM in order to test the following hypotheses: 1) the assumptions made with regard to the constancy of DFFM and HFFFM are systematically and significantly violated in patients with cirrhosis, even in the absence of overt fluid retention; 2) standard 2-CM techniques for assessing body composition provide inaccurate estimates in patients with cirrhosis; and 3) the extent of variation in the estimates of body composition variables obtained with the use of standard "reference" and "bedside" 2-CM techniques may preclude the development of meaningful population-specific correction factors or prediction equations.
SUBJECTS AND METHODS
Subjects
Twenty patients [10 men and 10 women;
Study outline
Data measurements in each patient were completed within 24 h. The patients attended the Royal Free Hospital, London, between 0700 and 0900 after fasting and refraining from smoking for 8 h. On arrival, the patients voided the last sample of a 24-h collection for estimation of urinary creatinine excretion. Clinical stability was ensured and anthropometric measures obtained. Baseline samples of saliva were collected, and a dosing solution containing deuterium was administered orally for the estimation of TBW. A snack providing a total of 200 kcal and 3 g protein was given to the patients before transfer to the MRC Dunn Clinical Nutrition Centre in Cambridge. Densitometry and DXA measurements were undertaken shortly after arrival; a small sandwich lunch was then provided. BIA and near-infrared interactance (NIRI) were undertaken 1 h later. Food and fluid intake were monitored until the collection of the 6-h postdose salivary sample for the deuterium dilution studies, after which intake was unrestricted.
The healthy subjects were recruited from the Cambridge area and attended the MRC Dunn Clinical Nutrition Centre directly from home. All reference measurements were undertaken by using the same protocol employed in the patient group; urinary creatinine data were not collected in these persons.
Assessment methods
Anthropometric measures
Height was measured to the nearest 0.5 cm, in bare feet, by using a freestanding stadiometer (Seca, Hamburg, Germany). Body weight (BWt) was measured to the nearest 0.1 kg with subjects in swimwear by using an electronic digital scale (Sauter Type E1210; Todd Scales, Newmarket, Suffolk, United Kingdom). Body mass index was calculated as BWt (in kg)/height2 (in m).
Skinfold anthropometric measures were undertaken by using the same steel tape measure and Holtain/Tanner-Whitehouse skinfold calipers (Holtain Ltd, Crymych, Dyfed, United Kingdom) in all subjects. Standardized techniques were used to measure the triceps, biceps, subscapular, and suprailiac skinfolds and midupper arm circumferences at set anatomical landmarks on the nondominant side of the body (24). Each measurement was made to the nearest 0.1 mm, and a mean result calculated from 3 or 4 readings. Bone-free, midupper arm muscle cross-sectional area was calculated (25).
Urinary creatinine excretion
Twenty-four-h urine samples were collected in plastic bottles containing boric acid preservative following a standard procedure. The completeness of the collection was assessed in 16 of the patients using para-amino benzoic acid (PABA; Laboratories for Applied Biology Ltd, London, United Kingdom); one 80-mg PABA tablet was taken at each of 3 regular meals during the 24-h collection period. Total 24-h urine volumes were measured, and aliquots were stored at –20 °C until assayed.
Urinary creatinine concentration was assessed calorimetrically on an automated Hitachi 911 instrument (Hitachi Europe Ltd., London, United Kingdom) by using its reaction with alkaline picrate (26); excretion rates were determined by extrapolation from the 24-h urinary volume. Recovery of PABA was measured calorimetrically by using its reaction with naphthylene ethylene diamine dihydrochloride; urine collections containing >205 g PABA (85% of the administered dose) were considered complete (27). Creatinine clearance was calculated from the serum and urinary creatinine concentrations and the 24-h urinary volume.
Deuterium dilution
A baseline salivary sample was collected with small balls of cotton wool, which were chewed and then expectorated into small sealable specimen jars. Care was taken to prevent fractionation with exogenous water. An accurately weighed amount of oral dosing solution was given to provide a nominal dose equivalent to 0.40 g deuterium oxide/kg BWt (99.9 atom percent excess; Sigma Chemical Company, Poole, Dorset, United Kingdom). The mouth was rinsed with an additional 150 mL tap water to minimize deuterium losses. A further salivary sample was collected 6 h after dosing.
The extracted saliva, which varied in volume from 4–8 mL, was centrifuged at 1207 x g for 10 min at 4° C, and the supernatant was transferred into sealed plastic tubes for storage at –20 °C until analyzed. Deuterium enrichment in salivary samples was measured, in duplicate, with Fourier transformed infrared spectroscopy (ATI Mattson Genesis Series FTIR; ATI, Cambridge, United Kingdom) as described previously (28). The difference in enrichment between the baseline and the 6-h postdose salivary sample was used to calculate deuterium dilution space. TBW was determined by using a correction factor of 1.04 for nonaqueous exchange (29, 30):
RESULTS
All 20 patients had stable, alcohol-related cirrhosis and had been abstinent from alcohol for a mean (range) 24.4 mo (3–76 mo). None had evidence of overt fluid retention either on clinical examination or on abdominal ultrasonography. Thirteen patients were classified as Child's grade A (7 men and 6 women) and 7 as grade B (3 men and 4 women). The median plasma creatinine concentration was 83 (57–100) µmol/L, and the median creatinine clearance was 84.1 (22.9–142.8) mL/min; 9 patients (3 men and 6 women) had creatinine clearances of <80 mL/min.
All subjects completed the study as per protocol. Urine collections were considered complete in 8 of the 16 patients; with a mean PABA recovery of 96% (85–106%). PABA excretion ranged from 24% to 82% in the remaining 8 patients.
Overall, there were few significant differences in anthropometric variables between the healthy controls and the patients with cirrhosis, although there was some evidence, in the male patients, of a reduction in muscle mass (Table 2).
View this table:
TABLE 2. Anthropometric variables in the patients with cirrhosis and in the pair-matched healthy control subjects1
No significant differences in mean body fat or in mean body FFM, derived from the 4-CM, were observed between the healthy controls and the patients with cirrhosis, either in the group as a whole (fat: 21.5 ± 8.7 compared with 22.7 ± 8.2 kg; FFM: 52.4 ± 11.7 compared with 50.1 ± 10.4 kg), or when separated by sex (Table 3).
View this table:
TABLE 3. Body composition measured by using the 4-component model in the patients with cirrhosis and in the pair-matched healthy control subjects1
The mean BMC in the patients with cirrhosis was significantly lower than in the healthy controls, in the group as a whole (2.14 ± 0.44 compared with 2.66 ± 0.49 kg; P < 0.001) and in both men (P < 0.001) and women (P < 0.001) when considered separately (Table 3). In consequence, significant intergroup differences in the derived values for TBM and the contribution of TBM to FFM were observed, because the same correction factor was used for both groups (Table 3).
The mean DFFM in the healthy controls, derived from the 4-CM, was comparable to the established reference mean of 1.1 kg/L (12) and to the values previously obtained in healthy men (1.1024 ± 0.0078 kg/L) and women (1.1003 ± 0.0066 kg/L) by using the same 4-CM (14) (Table 3). The mean DFFM in the patients with cirrhosis was significantly lower than that in the healthy controls, both in the group as a whole (1.091 ± 0.008 compared with 1.100 ± 0.006 kg/L; P < 0.001) and when separated by sex (Table 2).
The mean HFFFM in the healthy controls, derived from the 4-CM, was comparable to the reference value of 73.2% (32) and was not significantly different from the values previously obtained in healthy men (73.3 ± 2.2%) and women (74.5 ± 1.9%) by using the same 4-CM (14) (Table 3). Overall, the values for HFFFM were higher in the patients with cirrhosis than in the healthy controls, but the mean HFFFM did not differ significantly from control values either in the group as a whole or when separated by sex (Table 3).
The major proportion of the total variance observed in DFFM and HFFFM both in the healthy controls (89%) and in the patients with cirrhosis (94%) was accounted for by biological variance. The major factor responsible for the reduction in the mean DFFM in the patients with cirrhosis was the reduction in BMC, although the reductions observed in TBP and TBW in the male patients may also have contributed, even though the intergroup mean differences for these variables were not statistically significant.
Significant differences in the estimates of body fat and FFM obtained from the 4-CM and other established body composition techniques were observed in both the healthy controls and in the patients with cirrhosis, although the discrepancies in the patient population were greater (Table 4 and Table 5; Figure 1 and Figure 2).
View this table:
TABLE 4. Comparison of values for percentage body fat obtained by using various established techniques with the values obtained by using the 4-component model (4-CM) in the patients with cirrhosis and in the pair-matched healthy control subjects1
View this table:
TABLE 5. Comparison of values for fat-free mass obtained by using various established techniques with the values obtained by using the 4-component model (4-CM) in patients with cirrhosis and the pair-matched healthy control subjects1
FIGURE 1.. Mean (±SD) body fat assessed in 20 patients (10 men and 10 women) with cirrhosis () and in 20 (10 men and 10 women) pair-matched healthy control subjects () by using a 3- and a 4-component model (3-CM; 4-CM) of body composition (14) and a variety of other measurement techniques, including underwater weighing (UWW), deuterium dilution (Deuterium), dual-energy X-ray absorptiometry (DXA), skinfold thicknesses (SFT), bioelectrical impedance analysis (BIA; numbers represent references for the used BIA equation), near-infrared interactance (NIRI), and urinary creatinine excretion (Creatinine). *,**Bias and 95% limits of agreement were significantly different from the 4-CM (Bland and Altman): * P < 0.05, ** P < 0.01.
FIGURE 2.. Mean (±SD) fat-free mass (FFM) assessed in 20 patients (10 men and 10 women) with cirrhosis () and 20 (10 men and 10 women) pair-matched healthy control subjects () by using a 3- and a 4-component model (3-CM; 4-CM) of body composition (14) and a variety of other measurement techniques, including underwater weighing (UWW), deuterium dilution (Deuterium), dual-energy X-ray absorptiometry (DXA), skinfold thicknesses (SFT), bioelectrical impedance analysis (BIA; numbers represent references for the used BIA equation), near-infrared interactance (NIRI), and urinary creatinine excretion (Creatinine). *,**Bias and 95% limits of agreement were significantly different from the 4-CM (Bland and Altman): * P < 0.05, ** P < 0.01.
The urinary creatinine excretion provided the least accurate estimates of body fat and FFM in the patients with cirrhosis. Some improvements in the estimates were observed when data from patients in whom PABA excretion indicated an incomplete collection or in whom the creatinine clearance was <80 mL/min were excluded. Nevertheless, even with these exclusions, the technique still provided overall discrepancies exceeding 12% body fat and 10 kg FFM.
In men with cirrhosis, deuterium dilution provided estimates of body fat that were closest to those obtained with the 4-CM, whereas DXA provided the closest estimates of FFM; nevertheless, the 95% limits of agreement between the results exceeded 5% body fat and 3 kg FFM. The least accurate estimates of body fat and FFM were provided by creatinine excretion, the use of which resulted in discrepancies exceeding 14% body fat and 11 kg FFM (Tables 4 and 5; Figures 1 and 2).
In women with cirrhosis, BIA measured by using the Schloerb equation (38) provided estimates of body fat and FFM that were closest to those obtained with the 4-CM; nevertheless, the 95% limits of agreement between the results exceeded 5% body fat and 5 kg FFM. In these patients, the least accurate estimates of body fat and FFM were provided by creatinine excretion, the use of which resulted in discrepancies exceeding 25% body fat and 16 kg FFM.
The mean value for DFFM in the patients with cirrhosis (1.0907 kg/L) and the established mean for the density of fat (0.9007 kg/L) were applied, according to Archimedes principle (15), and simplified according to the format of the Siri equation (12), to provide a revised population-specific equation for estimating percentage fat from BD in this patient population, as follows:
DISCUSSION
The considerable difficulties associated with the assessment of body composition in patients with liver disease are well acknowledged (6-8). However, few attempts have been made, to date, to either use the more sophisticated multicomponent approach to provide direct and accurate measures of the FFM constituents (19-22) or to validate existing body composition methods appropriately.
The 4-CM used in the present study was considered to be an appropriate method for assessing body composition in patients with chronic liver disease and for evaluating alternative techniques. This model integrates measurements obtained from BWt, densitometry, deuterium dilution, and DXA to maximize the accuracy of estimates of body fat, water, mineral, and protein, with only slightly worse precision than that encountered when using the measurements individually (13, 14).
The 4-CM does not provide direct estimates of glycogen, urea, free-amino acids, and nucleic acids; these components are included primarily in the estimate of TBP because they have similar densities. However, in healthy persons, these additional components comprise only a small fraction of total body mass, of the order of 1 kg in a 70-kg man, and, in patients with cirrhosis, glycogen stores are depleted (44, 45) and protein synthesis is impaired (17). Thus, inclusion of these additional components should have no major effect on the estimates of TBP in this patient population.
No significant differences were observed in fat mass and FFM between the patients and the pair-matched healthy controls, but intergroup differences were observed in both the HFFFM and the DFFM. These discrepancies have considerable implications, because differences of 0.01 in HFFFM and of 0.01 kg/L in DFFM represent potential errors in the estimation of body fat of 1% and 5%, respectively.
The mean HFFFM value in the patients of 0.745 exceeded the mean in the healthy controls of 0.735; the greater variability observed in HFFFM values in the patients may have masked the significance of the intergroup difference. The upper limit of the HFFFM range in the patients exceeded that in the healthy controls by 4.5%. Thus, although none of the patients studied had evidence of fluid retention, some were clearly overhydrated. This agrees with the finding of McCullough et al (16) that abnormalities in fluid homeostasis and compartmentalization can be detected in patients with cirrhosis with no clinical or radiological evidence of fluid retention.
Although the increase in HFFFM may not be clinically relevant, it has implications for the assessment of body composition. FFM can be assessed from TBW by using a reference value for HFFFM of 0.732 (32). The mean HFFFM measured in this study of 0.745 could be substituted for the reference value only on a group basis, because the degree of variability in HFFFM precludes its use on an individual basis.
The mean DFFM in the patients with cirrhosis was significantly lower than that in the healthy controls partly because of both absolute and proportional reductions in BMC and TBM. This reduction in DFFM in the patients with cirrhosis is of clinical significance because it indicates a major risk for bone health. In these patients, bone formation is reduced and bone mineral metabolism is defective. In consequence, bone mass, particularly that of the lumbar spine, is often significantly reduced (11, 46, 47).
The reduction in DFFM also has implication for the assessment of body composition in this patient population. Techniques such as anthropometry and densitometry are based on a constant value for DFFM of 1.1 kg/L (12, 14), whereas the mean value in this patient population was 1.09 kg/L. Changing the value of the DFFM constant when applying these techniques to provide a population-specific equation improves the accuracy of mean values for body composition variables but does not provide more accurate estimates on an individual basis because of the relatively wide interindividual variation observed in the DFFM values.
The violations observed in the constant associations between body components in the patients with cirrhosis are reflected in the poor agreements observed between the assessments of body composition obtained by using the 4-CM and those obtained with the individual "reference" techniques, densitometry, deuterium dilution, and DXA. Deuterium dilution, which adopts assumptions with regard to the HFFFM, provided the least inaccurate estimates of body composition in the patients with cirrhosis, but the overall discrepancies still amounted to 5% body fat and 3.5 kg FFM.
DXA technology, which also adopts assumptions with regard to the HFFFM, provided estimates of body composition in the patients with cirrhosis with discrepancies up to 6% body fat and 3.7 kg FFM. This technique has been used to provide reference body-composition data in patients with cirrhosis (19-21, 46), but a more detailed evaluation of the technique, in particular its ability to detect changes in body composition in response to treatment or nutritional intervention, is required before its use as a single technique for body composition analysis can be recommended (48).
The body-composition data obtained in the patients with cirrhosis with the "bedside" techniques of anthropometry, BIA, NIRI, and creatinine excretion were, as expected, even less accurate than those provided by the individual "reference" techniques. None provided estimates that confidently predicted body fat to within 11% or FFM to within 7.5 kg of the values determined with the use of the 4-CM. These techniques, with the exception of NIRI, are used routinely in clinical practice to assess and monitor nutritional status in patients with liver disease and in the research setting to provide reference variables for measures such as energy expenditure. None has, however, been validated for these purposes.
The accuracy of the results obtained with BIA is critically dependent on the equation used to process the data. Nevertheless, even the most favorable provided estimates that were discrepant by 11% for body fat and 7.5 kg for FFM, compared with the 4-CM. The accuracy of BIA in the presence of significant fluid retention is likely to be significantly lower (20, 49).
The body-composition data based on urinary creatinine excretion were the least accurate of all the techniques investigated, providing estimates that were discrepant by 21% for body fat and 14 kg for FFM, compared with the 4-CM. Reduced serum creatinine concentrations have been reported in patients with cirrhosis and are generally attributed to low protein intake, impaired hepatic synthesis, a reduction in muscle mass (50), or an increase in renal tubular secretion (51, 52). Pirlich et al (53) recommended that creatinine excretion should not be used to assess body composition variables in patients with cirrhosis, unless the creatinine clearance exceeded 80 mL/min. In addition, the completeness of the urine collection should always be validated (54). However, even with these impositions, the technique still provided overall discrepancies exceeding 12% body fat and 10 kg FFM in the patients with cirrhosis.
In theory, values for the comparison bias could be used to correct the estimates of body composition derived by using these alternatives techniques (42). However, although group mean values would be valid after such a correction, the 95% limits of agreement would remain substantial, casting doubt on the utility of the corrected data in individual patients.
In conclusion, the assumptions made regarding the constancy of DFFM and HFFFM were systematically and significantly violated in patients with cirrhosis, even in the absence of overt fluid retention. Consequently, the standard methods for assessing body composition not only provide inaccurate estimates in this patient population, but the extent of the variation precludes the development of meaningful population-specific correction factors or prediction equations. It is therefore recommended that, for research purposes or where accurate and precise body-composition data are required, they should be obtained by using a 4-CM or other appropriate multicomponent technique. The underwater weighing procedure is difficult to access, but the less demanding technique of air-displacement plethysmography may be usefully substituted to obtain densitometry data (55). If these facilities are unavailable, then the considerable limitations of the alternative "reference" and "bedside" techniques must be recognized and taken into account when assessing and interpreting the data acquired. Finally, the findings of the present study indicate a compelling need to address the problems of body-composition assessment in patients with cirrhosis, especially in the clinical setting.
ACKNOWLEDGMENTS
We thank the patients for taking part in the study and for their enthusiasm and encouragement and Dr Richard Morris for statistical advice.
MYM and ME originally conceived the study. MYM, ME, AMM, and NJF designed the study. MYM and AMM selected and recruited the patients. AMM and NJF coordinated and undertook the study. GJ undertook the deuterium and PABA assays. NJF, MYM, and AMM undertook the statistical analysis. AMM wrote the first draft of the paper, which was then revised by MYM, NJF, and ME. All authors contributed to the subsequent revision of the manuscript. None of the authors had a personal or financial conflict of interest.
REFERENCES