Archives of Physical Medicine and Rehabilitation
Volume 84, Issue 6 , Pages 854-861, June 2003

The use of bioelectric impedance analysis to measure fluid compartments in subjects with chronic paraplegia1 ☆

  • Andrea C Buchholz, RD, PhD(c)

      Affiliations

    • Department of Nutritional Sciences, University of Toronto, Toronto, ON Canada
    • The Research Institute, The Hospital for Sick Children, Toronto, ON, Canada
    • ,
  • Colleen F McGillivray, MD, FRCP(c)

      Affiliations

    • Department of Medicine, University of Toronto, Toronto, ON Canada
    • the Toronto Rehabilitation Institute, Toronto, ON, Canada
    • ,
  • Paul B Pencharz, MB, ChB, PhD, FRCP(c)

      Affiliations

    • Corresponding Author InformationReprint requests to Paul B. Pencharz, MB, ChB, PhD, FRCP(C), Rm 8263, Div of Gastroenterology and Nutrition, The Hospital for Sick Children, 555 University Ave, Toronto, ON M5G 1X8, Canada
    • Department of Nutritional Sciences, University of Toronto, Toronto, ON Canada
    • Department of Paediatrics, University of Toronto, Toronto, ON Canada
    • The Research Institute, The Hospital for Sick Children, Toronto, ON, Canada

Article Outline

Abstract 

Buchholz AC, McGillivray CF, Pencharz PB. The use of bioelectric impedance analysis to measure fluid compartments in subjects with chronic paraplegia.

Objectives:

To determine the sensitivity and specificity of body mass index (BMI) as a surrogate marker of obesity in individuals with chronic paraplegia and to validate bioelectric impedance analysis (BIA) as a method of measuring body composition in this group.

Design:

Cross-sectional study.

Setting:

University hospital.

Participants:

Convenience sample of 31 subjects with paraplegia (19 men, 12 women; mean age, 34.2±8.8y) and 62 able-bodied control subjects (30 men, 32 women; mean age, 28.6±7.2y).

Interventions:

Not applicable.

Main Outcome Measures:

Total-body water (TBW) by deuterium dilution; extracellular water (ECW) by corrected bromide space. Fat-free mass (FFM)=TBW/.732; fat mass (FM)=weight−FFM. Single-frequency whole-body and segmental BIA, and multifrequency whole-body BIA.

Results:

BMI had 100% specificity and 20% sensitivity in distinguishing obese from nonobese subjects with paraplegia. TBW was predicted by using the equation: TBW (inL)=2.11−0.1age+3.45sex+.34wt+.28(ht2/R)−.086sex×wt(r2=.95, standard error of the estimate [SEE]=1.86L, P<.0001). This equation had 81.8% specificity and 68.4% sensitivity. ECW was predicted by using the equation: ECW (in L)=−.025+1.03sex+.187wt+.0041(ht2/Xc) −.033sex×wt (r2=.75, SEE=1.62L, P<.0001). Multifrequency BIA offered no greater prediction of TBW or ECW than single-frequency BIA.

Conclusions:

BMI has excellent specificity but poor sensitivity in distinguishing obese from nonobese individuals with paraplegia. TBW (and therefore FFM and FM) and ECW can be reasonably well predicted by using single-frequency BIA.

Keywords:  Body composition, Body water, Electric impedance, Obesity, Paraplegia, Rehabilitation

 

ALTERATIONS IN BODY COMPOSITION, specifically increased fat mass (FM) and decreased fat-free mass (FFM), have been associated with numerous health conditions. In the paraplegic population, increased FM (obesity) is associated with glucose intolerance and insulin resistance,1, 2 hyperlipidemia,3 coronary artery disease,4 increased pain,5 and compromised mobility.6 Spinal cord injury (SCI) is also accompanied by decreased FFM, in the form of bone mineral content and lean soft tissue.7, 8, 9, 10, 11, 12 Total-body water (TBW) is quantitatively the most important component of lean soft tissue and can be further subdivided into intracellular water (ICW) and extracellular water (ECW). ICW is a proxy for body cell mass (skeletal muscle and organs) and has been found to be depressed in SCI8 because of disuse atrophy. ECW, comprising interstitial fluid, plasma, and transcellular fluid, has been found to be increased in SCI7; this may be because of edema secondary to previous or chronic thromboembolism and paralysis of the lower limbs with an ineffective calf pump.13, 14, 15

Measuring body composition is therefore important for identifying patients at risk for the sequelae outlined earlier. However, well-recognized methods of measuring body composition, such as hydrodensitometry, dual energy x-ray absorptiometry, isotope dilution, or corrected bromide space are either not readily available at the clinical level and/or require considerable technical expertise. One method being used with increasing frequency and requiring minimal technical expertise is bioelectric impedance analysis (BIA). BIA measures the resistance of body tissues to an electric current. From resistance measures, prediction equations to estimate TBW (and subsequently FFM and/or FM) may be developed, as has been done in clinical populations including Duchenne’s muscular dystrophy,16 amyotrophic lateral sclerosis,17 human immunodeficiency virus,18 chronic obstructive pulmonary disease,19 and acromegaly.20 Although BIA has been used in the population with SCI,21, 22 it has not been validated, nor have SCI-specific equations been developed.

Increased FM is usually accompanied by increased body weight. In the absence of a clinically available and validated tool for measuring body composition, indices of relative weight, such as body mass index (BMI; in kg/m2), are used to distinguish those who are obese from those who are not. A BMI ≥30kg/m2 is considered an indicator of obesity.23 The main assumption of BMI guidelines is that body mass adjusted for stature squared is closely associated with body fatness and consequent morbidity and mortality.24 Several studies in the SCI literature report mean BMIs between 21.4 to 24.8kg/m2 and body weights within the normal range.9, 12, 22 However, percentage FM (% FM) in studies of men with SCI ranges from 23% to 30%, indicative of obesity.8, 9, 10, 12, 22, 25, 26, 27 Despite this, BMI continues to be used at the clinical level and to be reported in the SCI literature.28, 29, 30, 31 It has been suggested that the validity of BMI as a surrogate measure of obesity in SCI be determined.32

The objectives of this study were 2-fold. First, we wanted to determine the sensitivity and specificity of BMI as a surrogate measure of obesity in a group of men and women with chronic paraplegia. Second, we wanted to validate BIA against deuterium dilution (for TBW and, by calculation, FM) and corrected bromide space (for ECW), to develop predictive equations that may be used at the clinical level to estimate body composition in persons with chronic paraplegia.

Back to Article Outline

Methods 

Healthy subjects with paraplegia were recruited from the Lyndhurst Centre of the Toronto Rehabilitation Institute, Ontario Wheelchair Sports Association, Ontario March of Dimes, Canadian Paraplegic Association, and the Spina Bifida and Hydrocephalus Association of Toronto, ON. Healthy able-bodied subjects were recruited from the University of Toronto, Ryerson University, and staff of The Hospital for Sick Children, in Toronto. Subjects were recruited using flyers and pamphlets. Data collection began in April 1999 and was completed in August 2001. Subjects were group matched on the basis of BMI. All subjects completed a health screening questionnaire: none reported having active decubitus ulcers or a history of diabetes, Crohn’s disease, renal disease, or heart disease. Subjects were euthyroid on the basis of thyroid stimulating hormone, triiodothyronine, and free thyroxine levels (data not shown). Women were in the self-reported follicular phase of menses. The study was approved by the Research Ethics Boards of The Hospital for Sick Children and The Toronto Rehabilitation Institute.

Procedures 

Studies were performed during a 1-day visit to the Clinical Investigation Unit of The Hospital for Sick Children. Subjects were instructed not to exercise or to consume alcohol or caffeine for the 24 hours preceding the study day. They arrived in the morning after a 12-hour fast and provided informed consent. All measures were conducted by the same investigator (ACB), with subjects wearing light clothing or a hospital gown and no shoes.

Subjects were first asked to empty their bladders. Weight was then measured to the nearest 0.1kg on a beam balance scalea for the control subjects and on a digital wheelchair scaleb for subjects with paraplegia. The coefficient of variation (CV) between the 2 instruments was determined in a subsample of 6 able-bodied control subjects and was .36%±.15%. Height was measured to the nearest 0.1cm with a wall-mounted stadiometerc for control subjects and on an adult-sized Plexiglas length board (made specifically for the study by the Medical Engineering Department of The Hospital for Sick Children, following specifications33) for subjects with paraplegia. Subjects were asked to transfer from their wheelchairs to a bed and on to the length board. With the subject’s head resting against the immovable headboard, legs outstretched and feet in dorsiflexion, the movable footboard was pressed against the heels. Subjects looked at the ceiling during the measurement, while the investigator ensured that hips were straight and centered on the board. The CV between the stadiometer and the length board was .78%±.23%.

Whole-body and segmental bioelectric impedance measurements were taken on the right side of the body with gel electrodes by using a fixed (50kHz) frequency BIAd at 800mÅ and standardized procedures.34, 35 Subjects were supine, with the head of the bed elevated about 0° to 10°. All sites were cleaned with an alcohol swab before attachment of electrodes and arms and legs were abducted 30° from midline during all measures. Electrodes were positioned in the middle of the dorsal surfaces of the hands and feet proximal to the metacarpophalangeal and metatarsophalangeal joints, respectively, and also medially between the distal prominences of the radius and the ulna and between the medial and lateral malleoli of the ankle.

For measurements of the leg, 1 pair of electrodes was placed at the conventional locations on the foot and ankle and the other pair on the anterior midline of the proximal thigh, with the signal-detecting electrode in the same plane as the gluteal crease and the signal-introducing electrode 5cm proximal to the detecting electrode. Trunk impedance was measured by placing 1 pair of electrodes on the proximal thigh at the same locations used for measurement of the leg but with the introducing electrode 5cm distal to the detecting electrode. The other pair was placed with the detecting electrode over the sternal notch and the introducing electrode on the anterior midline of the neck 5cm cranial to the detecting electrode. The impedance of the arm was measured by attaching 1 pair of electrodes at the conventional locations on the hand and wrist and the other pair on the anterior surface of the shoulder with the measuring electrode over the midpoint of a line between the acromial process and the axilla and the introducing electrode 5cm medial to this midpoint.

Length of body segments was measured with a flexible measuring tape between the middle of the 2 inside electrodes for each of the legs, arms, and trunk. Multifrequency BIAe was measured in subjects with paraplegia only. Whole-body measures were performed by using the electrode configuration outlined above at each of 5, 50, and 200kHz. The calibration of both analyzers was checked daily against a 500Ω precision resistor; measured resistance for the fixed frequency instrument was consistently within 1.2% and for the multifrequency instrument within 0.4%.

After taking a baseline blood sample (15mL, taken with a heparinized syringe), each subject was given an oral dose of water labeled with deuteriumf (2H2O) for the measurement of TBW and sodium bromideg (NaBr) for the measurement of ECW, as previously described.36 Briefly, the doses were .25g of 99.9 atom percent 2H2O/kg estimated TBW (assuming 60% of body weight is TBW) and 1.0mL of 3% NaBr/kg body weight. The container was then rinsed with about 15mL of deionized water, which the subject subsequently drank to ensure the entire dose was consumed. A plateau blood sample (15mL) was obtained 3 hours after administration of the 2H2O and NaBr.37, 38 Subjects continued fasting during the equilibration period. Blood samples were centrifugedh at −4°C at 1200g for 10 minutes and plasmas were stored at −20°C until analyses.

Body composition 

Plasma samples were analyzed for their 2H2O content by using a continuous flow isotope ratio mass spectrometer,i after equilibration with hydrogen gas.39 TBW (in liters) was calculated as ([(dose×99.9)/20]×[18.02/APE]×10−3)/1.04, where dose is the dose of 2H2O in grams, 99.9 is the atom percentage of 2H2O, 20 is the molecular weight of 2H2O, 18.02 is the molecular weight of unlabeled water, APE is atom percentage excess (APplateau−APbaseline), and 1.04 is the correction for hydrogen dilution space. FFM (in kilograms) was calculated as TBW divided by .732, by using the lean tissue hydration constant of Pace and Rathbun,40 and FM (in kilograms) as weight minus FFM. ECW (in liters) was determined from corrected bromide space from plasma samples by means of the bromide dilution technique.38 Bromide concentration in the ECW space was determined from plasma samples41 by neutron activation of the stable 79Br to 80Br and using the following equation: ECW (in L)=(Br dose/[plasma enrichment at 3h corrected for plasma enrichment at baseline])×.90×.95×.94, where .90 is the correction factor for nonextracellular bromide distribution, .95 is the Donnan equilibrium factor, and .94 is the correction for water in plasma. Body cell mass (BCM, in kg) was determined as (TBW−ECW)/.732.

% FM cutoff points for determination of validity of BMI 

Body mass index was calculated as weight (kg)/height (m2). The standard BMI cutoff value of ≥30kg/m2, as used by the World Health Organization (WHO), was used to identify obese subjects.23 Sex-specific % FM cutoff points were used to classify subjects as obese (men: 18–40y, >22% FM; 41–60y, >25% FM; women: 18–40y, >35% FM; 41–60y, >38% FM).42 This allowed for the assessment of the sensitivity and specificity of BMI for distinguishing between obese versus nonobese subjects.

Statistics 

Data are presented as mean ± standard deviation (SD). The SAS programj was used for all computations. Results were statistically significant at P less than .05. Nonnormal data were log transformed. A chi-square test was performed to determine whether there were differences in sex distribution between the 2 study groups. Differences between the 2 were determined by using Student t tests and, where warranted, P values for unequal variances. A 2-way analysis of variance for unbalanced designs was done to determine whether there was a significant group by sex interaction for % FM. Sensitivity of BMI was determined as (no. true positive/[no. true positive+no. false negative])×100% and specificity as (no. true negative/[no. false positive+no. true negative])×100%. The receiver operator characteristic (ROC) curve was then used to determine a more relevant BMI cutoff. The Pearson correlation coefficient was used to quantify the univariate association between TBW or ECW and selected predictor variables; the coefficient (r) is presented as r2, the percentage variance in TBW (or ECW) explained by the predictor variables. These associations were further evaluated by using the multivariate technique of all possible regressions and Mallow’s Cp statistic. Possible predictors of TBW and ECW in both groups included age, sex, weight, height, R (resistance at 50kHz, in Ω), Xc (reactance at 50kHz, in Ω), ht2/R (height in cm2/R), (lengtharm)2/Rarm, (lengthleg)2/Rleg, and (lengthtrunk)2/Rtrunk. For the group with paraplegia only, additional predictors of TBW included R200 (resistance at 200kHz, in Ω) and ht2/R200, and Xc200 (reactance at 200kHz, in Ω), and of ECW included R5 (resistance at 5kHz, in Ω), ht2/R5, and Xc5 (reactance at 5kHz, in Ω). Models were tested for interaction and for multicolinearity (the latter using the variance inflation factor). Bland and Altman analyses43 were performed to determine bias, and paired t tests were used to determine significant differences, between predicted and measured TBW or ECW.

Back to Article Outline

Results 

Participants 

Thirty-one adults with paraplegia and 62 able-bodied adults volunteered to participate in the study. There were 19 men and 12 women in the group with paraplegia and 30 men and 32 women in the control group (χ2=1.38, P=.2400). The most common cause of paraplegia was motor vehicle crash (n=11), followed by spina bifida (n=7), hemorrhage (n=4), and falls (n=3). The remaining etiologies were mixed and included transverse myelitis (n=2), gunshot wound (n=1), scuba diving mishap (n=1), bacterial infection (n=1), and von Hippel Lindau syndrome (n=1). The mean number of years since onset of paraplegia was 13.8±11.8 years. Twenty-one subjects had complete and 10 had incomplete lesions; 7 had spinal hardware, and 24 did not. There were no differences in body composition between these subgroups, and so data were collapsed.

Body composition and validity of BMI 

Ages and body composition of both groups are shown in table 1. As expected, the group with paraplegia had significantly lower TBW, FFM, and BCM, but greater FM, % ECW, and ECW:TBW than controls. The % FM of the subjects by group and sex is shown in figure 1. The main effects of group and sex were significant (both P<.0001), but there was no group by sex interaction, indicating that the difference in % FM between men and women with paraplegia did not differ from that of male and female control subjects (F=.07, P=.7889).

Table 1. Age, Body Composition, and BIA Parameters in 62∗ Control Subjects and 31∗ Subjects With Paraplegia
ParameterControlParaplegicP
Mean ± SDRangeMean ± SDRange
Age (y)28.6±7.219–5534.2±8.820–57.0012
Height/length (cm)170±10.3144–202162±14.6128–183.0160
Weight (kg)66.6±11.241.8–98.064.3±16.242.9–100.3035
BMI (kg/m2)22.9±2.019.3–27.624.6±6.415.7–44.2.3468
TBW (L)37.3±7.922.3–57.932.2±8.021.7–50.3.0024
TBW (% body weight)55.6±5.144.8–68.250.6±6.835.4–63.5.0001
FFM† (kg)51.0±10.830.5–79.144.0±11.029.6–68.7.0024
% FFM76.0±7.061.2–93.269.2±9.248.4–86.8.0001
FM (kg)15.8±4.84.9–26.820.2±9.08.0–47.5.0142
% FM24.0±7.06.8–38.830.8±9.213.2–51.6.0001
ECW (L)14.4±3.17.4–23.515.1±4.07.9–28.3.3368
ECW (% body weight)21.4±2.615.4–28.723.8±4.316.8–34.4.0069
ECW:TBW0.38±0.050.30–0.570.48±0.080.33–0.69<.0001
BCM‡ (kg)31.6±7.516.7–51.023.1±7.99.3–44.5<.0001
BCM (% body weight)46.8±6.528.4–61.236.1±8.218.5–51.2<.0001

NOTE. All values are mean ± SD or range.

∗ n=60 control subjects for ECW and 61 for TBW; n=30 subjects with paraplegia for ECW.

† FFM=TBW/.732.

‡ BCM=[(TBW−ECW)/.732].

  • View full-size image.
  • Fig 1. 

    % FM of all subjects. Men with paraplegia (P) had higher % FM than male controls (C) (26.8±7.3 vs 19.1±4.8, P<.0001), and women with paraplegia had higher % FM than female controls (37.1±8.6 vs 28.7±5.4, P<.0001). There was no group by sex interaction (F=.07, P=.7889).

The mean BMI of each group was within the normal-to-overweight range and did not differ significantly. However, the actual prevalence of obesity (as determined by deuterium dilution) in the group with paraplegia was 64.5% and in the control group was 23.0%. The specificity of BMI was 100% in both groups. That is, all nonobese subjects were correctly identified with BMI less than 30kg/m2; there were no false positives. We could not calculate the sensitivity of BMI in the control group because no subject had a BMI ≥30kg/m2. However, only 20% of obese subjects with paraplegia were correctly identified with a BMI ≥30kg/m2, indicating poor sensitivity; there were many false negatives. The mean BMI of these obese subjects was 27.6±6.0kg/m2 (range, 20.5–44.2kg/m2). Results of the ROC curve indicated that moving the obesity cutoff from ≥30 to ≥25kg/m2 improved the sensitivity to 60% without compromising specificity, which remained at 100%.

Bioelectric impedance analysis 

Whole-body and segmental BIA parameters are shown in table 2. Resistance for each of the whole body, leg, and trunk was significantly higher in the group with paraplegia, indicating less TBW (and therefore FFM) in these body segments. Conversely, the resistance of the arm was lower, indicating greater TBW in this body segment than the control group. Reactance for the whole body and for the leg was significantly lower in the group with paraplegia, suggesting lower BCM in these body segments, whereas reactance of the arm was higher, suggesting greater BCM in this body segment. There was no difference in the reactance of the trunk between the 2 groups.

Table 2. Whole-Body and Segmental BIA Parameters in 62∗ Control Subjects and 31∗ Subjects With Paraplegia
ControlParaplegicP
Mean ± SDRangeMean ± SDRange
Whole body
R (Ω)519±74.4352–683564±103390–825.0375
Xc (Ω)52.0±8.337.0–76.044.1±8.929.0–66.7<.0001
Segmental
Leg R246±31.4179–338342±97.7192–681<.0001
Leg Xc31.4±5.421.0–48.318.0±8.66.0–38.6<.0001
Leg length (cm)73.8±5.564–8870.1±9.146–84.0450
Arm R274±52.7166–361248±45.8180–357.0222
Arm Xc27.6±4.220–3831.8±5.019–40.7.0001
Arm length (cm)51.9±3.642.5–61.051.8±3.145–60.8951
Trunk R83.7±16.353.7–132115±24.574.3–173<.0001
Trunk Xc15.4±6.38–4214.4±5.27.3–33.4368
Trunk-length (cm)56.3±6.444.0–70.053.5±5.545.0–65.0.0423

NOTE. All values are mean ± SD or range.

∗ n=60 control subjects for arm reactance; n=30 subjects with paraplegia for whole-body resistance and reactance.

Correlations between ECW, TBW, and predictor variables are shown in table 3. The single best predictor of ECW in the group with paraplegia was ht2/R (r2=.58, standard error of the estimate [SEE]=2.65L, P<.0001) and in the control group it was weight (r2=.70, SEE=1.69L, P<.0001). Low-frequency BIA offered no greater prediction than BIA at 50kHz in the group with paraplegia: r2 between ECW and R5 was .45 (P<.001), ht2/R5 was .58 (P<.0001), and Xc5 was .012 (not significant). The technique of all possible regressions indicated that the best equation for predicting ECW in the control group was:

(r2=.75, SEE=1.62L, P<.0001), where sex is 0 for men and 1 for women, weight is in kilograms, height is in centimeters, and Xc is in ohms. When this equation was applied to the group with paraplegia, predicted ECW correlated strongly with measured ECW (r2=.66, SEE=2.38L, P<.0001), as shown in figure 2A. Bland and Altman analysis revealed no significant bias between predicted and measured ECW (P=.3276, fig 2B), and a paired t test showed only a trend toward a significant difference between predicted and measured ECW (−.86±2.36L, P=.0606).

Table 3. Pearson Correlation Coefficients (r2) Between TBW, ECW, and Selected Predictor Variables in 62 Control Subjects and 31 Subjects With Paraplegia
VariableECWTBW
ControlParaplegicControlParaplegic
Age.0003.02.0006.10
Weight.70∗.54∗.80∗.70∗
Height.57∗.39†.72∗.57∗
R.41∗.47∗.63∗.64∗
ht2/R.64∗.58∗.88∗.82∗
Xc.10‡.09.04.007
ht2/Xc.42∗.51∗.41∗.33∥
(lengthleg)2/Rleg.33∗.48∗.47∗.65∗
(lengtharm)2/Rarm.56∗.31∥.86∗.57∗
(lengthtrunk)2/Rtrunk.35∗.20‡.57∗.41†

∗ P<.0001.

† P<.001.

‡ P<.05.

∥ P<.01.

  • View full-size image.
  • Fig 2. 

    Relationship between predicted and measured ECW in 60 control subjects (squares, solid trendline) and 30 subjects with paraplegia (triangles, dotted trendline). ECW was predicted by using the equation: ECW (in L)=−.025+1.03sex+.187wt+.0041(ht2/Xc)−.033sex×wt. (A) r2 between predicted and measured ECW for controls was .75 (SEE=1.62L); r2 for subjects with paraplegia was .66 (SEE=2.38L) (both P<.0001). Slopes did not differ significantly (P=.4797). (B) Bland and Altman43 analysis in the group with paraplegia revealed no significant bias (r2=.04, P=.3276). Horizontal line represents mean difference between predicted and measured ECW (−.86±2.36L, P=.0606). Abbreviations: CBS, corrected bromide space.

The single best predictor of TBW in both groups was ht2/R (paraplegia: r2=.82, SEE=3.57L; control: r2=.88, SEE=2.74L, both P<.0001) (table 3). High-frequency BIA offered no greater prediction than BIA at 50kHz in the group with paraplegia: r2 between TBW and R200 was .58, ht2/R200 was .79 (both P<.0001), and Xc200 was .40 (P<.001). We subsequently developed 3 prediction equations for TBW, all based on single-frequency BIA. The technique of all possible regressions indicated that the best equation in the control group was:

1
(r2=.95, SEE=1.86L, P<.0001), where age is in years, sex is 0 for men and 1 for women, weight is in kilograms, height is in centimeters, and R is in ohms. When equation 1 was applied to the group with paraplegia, predicted and measured TBW were strongly correlated (r2=.91, SEE=2.48L, P<.0001), as shown in figure 3A. However, Bland and Altman43 analysis revealed a weak but significant bias (P=.0356, fig 3B), confirmed by a significant difference in slopes (P=.0041), and a paired t test showing predicted TBW to overestimate measured TBW by 1.70±2.84L (P=.0026). This resulted in an underestimation of FM by 2.0kg, or 2.7% of body weight. Despite this, the equation demonstrated greater sensitivity (68.4%) than BMI in distinguishing obese from nonobese subjects with paraplegia, although at the expense of specificity (81.8%).

  • View full-size image.
  • Fig 3. 

    Relationship between measured and predicted TBW in 61 control subjects (squares, solid trendline) and 31 subjects with paraplegia (triangles, dotted trendline). TBW was predicted by using equation 1: TBW (in L)=2.11−0.1age+3.45sex+.34wt+.28(ht2/R)−.086sex×wt. (A) r2 between predicted and measured TBW for controls was .95 (SEE=1.86L); r2 for subjects with paraplegia was .91 (SEE=2.48L) (both P<.0001). Slopes differed significantly (P=.0041). (B) Bland and Altman43 analysis in the group with paraplegia revealed significant bias (r2=.15, P=.0356). Horizontal line represents mean difference between predicted and measured TBW (1.70±2.84L, P=.0026).

We also developed 2 additional prediction equations for TBW in the group with paraplegia, as follows:

2
3
These equations performed well—equation 2: r2 was .96 and SEE was 1.84L; equation 3: r2 was .95 and SEE was 1.98L (both P<.0001). Bland and Altman43 analyses indicated no significant bias of either equation in predicting measured TBW—equation 2: r2 was .0112 (P=.5841); and equation 3: r2 was .0078 (P=.6489). Paired t tests indicated no significant differences between predicted and measured TBW—equation 2: mean difference was −.017±1.67L (P=.9571); and equation 3: mean difference was .21±1.79L (P=.5246). Equation 2 was devised by using all predictor variables in the regression model, whereas equation 3 was devised by using predictor variables independent of height. We did this recognizing that it may be cumbersome to measure length (height) in persons with paraplegia.

Back to Article Outline

Discussion 

The major findings of this study are (1) relative to deuterium dilution, the widely used BMI cutoff for obesity (≥30kg/m2) has excellent specificity, but poor sensitivity, in distinguishing obese from nonobese persons with paraplegia; and (2) ECW and TBW can be reasonably well predicted in persons with chronic paraplegia using single-frequency BIA and the following validated equations: ECW (in L)=−.025+1.03sex+.187wt+.0041(ht2/Xc)−.033sex×wt and TBW (in L)=2.11−0.1age+3.45sex+.34wt+.28(ht2/R)−.086sex×wt. Recognizing that the latter equation significantly overestimated measured TBW, we also devised 2 additional equations in the group with paraplegia (one which includes measurement of height; one which does not) and recommend that they be cross-validated in an independent sample of men and women with chronic paraplegia.

Validity of BMI 

In our study, all subjects who were truly nonobese by deuterium dilution were correctly identified with a BMI of less than 30kg/m2, indicating excellent specificity (no false positives) and confirming other findings in the literature.44, 45, 46 However, only 20% of obese subjects with paraplegia were correctly identified with a BMI ≥30kg/m2, indicating poor sensitivity (many false negatives). This is lower than published sensitivity values in able-bodied populations, which range from 48% to 66%.44, 45, 46 These results suggest that individuals with paraplegia and a BMI ≥30kg/m2 are likely truly obese, whereas no conclusions about body fatness can be made for those with a BMI less than 30kg/m2. The mean BMI of obese subjects with paraplegia was 27.6±6.0kg/m2 (range, 20.5–44.2kg/m2), suggesting that the WHO23 cutoff of 30kg/m2, intended for use in able-bodied individuals, may result in a substantial number of obese persons with paraplegia being missed on screening. Moving the obesity cutoff to ≥25kg/m2 improved sensitivity to 60%, without compromising specificity. Given that this cutoff was devised in a small number of subjects, we recommend that it be considered preliminary and be used with caution.

Prediction of fluid spaces by single- and multiple-frequency BIA 

The prediction equation for ECW performed well. Predicted and measured ECW correlated strongly and significantly, and there was no significant bias or difference between the 2. This is the first report to have validated a clinical tool to measure ECW in this population, which may be helpful in identifying and monitoring edema secondary to thromboembolism.13, 14, 15

We developed 3 TBW prediction equations, one of which was cross-validated in the group with paraplegia, and 2 of which require cross-validation. From these equations, FFM (TBW/.732) and FM (weight−FFM) may be determined. TBW equation 1 was developed in the control group and tested in the group with paraplegia. It performed reasonably well; there was a strong and significant correlation between predicted and measured TBW. Further, the sensitivity (68.4%) of equation 1 in distinguishing obese from nonobese persons with paraplegia was greater than that of the BMI cutoff of ≥30kg/m2 (20.0%); this was, however, at the expense of specificity, which was lower with equation 1 (81.8%) than with BMI (100%). TBW equation 1 significantly overpredicted measured TBW, similar to findings reported by other investigators.47, 48, 49 This may be because of the expanded ECW seen in paraplegia. An expansion in ECW, which has excellent conductive properties, decreases resistance. Because resistance (R) and TBW are inversely proportional, TBW is subsequently overestimated.

Ideally, we would have randomly divided a large number of subjects with paraplegia into 2 groups, developed an equation in 1 group, and cross-validated it in the other. Our sample size (n=31) did not allow for this. Instead, we devised equation 2 for the group with paraplegia. It performed well and showed no significant bias in predicting measured TBW. Both equations 1 and 2 require the use of a length board to determine height in persons with paraplegia. This board can be made locally following specifications.33 Using patients’ self-reported height is not recommended because over- or underestimation of height by 2.5cm has been shown to cause an error of 1.0L of TBW.50 In our study, this results in an error of 1.37kg FM, or 2.1% of body weight. Recognizing that not all centers will have the means to make an adult-sized length board, we devised equation 3, which includes resistance and length of the leg and trunk. This equation performed well and showed no significant bias in predicting measured TBW, but, as with equation 2, it requires cross-validation.

A difficulty with quantification of body composition by BIA is that most body compartments of interest, including TBW, ECW, ICW, and FFM, are themselves intercorrelated.51 We used multifrequency BIA to distinguish between TBW and ECW. This technique is based on the principle that resistance depends on the frequency of the current applied. Theoretically, at low frequencies, predominantly ECW is measured because of the capacitive effect of the cell membranes separating ICW from ECW. At high frequencies, the membranes are permeable to the current so that resistance reflects both the intra- and extracellular spaces, that is, TBW.52, 53 In our study, the prediction of ECW did not improve with low-frequency measurements, nor did the prediction of TBW improve with high-frequency measurements, similar to findings of others.53, 54, 55, 56, 57, 58, 59 Although it presents many practical advantages over deuterium and bromide dilution, we believe the clinical utility of using multifrequency BIA to distinguish TBW from ECW has yet to be clearly shown.

Segmental BIA 

BIA makes the assumption that the body is a conducting cylinder of uniform length and cross-sectional area. This is not entirely correct. The shape of the body more closely resembles a series of 5 cylinders (2 arms, 2 legs, trunk). Because resistance is inversely proportional to cross-sectional area, the upper and lower extremities (which have the smallest cross-sectional area) will have the most influence on whole body resistance measurements.60 Baumgartner et al35 found the accuracy of the prediction of FFM from (lengtharm)2/Rarm to be only marginally less than that obtained by using the whole-body measurements and concluded that measurements of the resistance and length of the arm can be used in place of the whole-body method. We also found the correlation between (lengtharm)2/Rarm and TBW in the control group to be excellent and very close to that of ht2/R. However, the use of(lengtharm)2/Rarm assumes that the composition of the arm is representative of the rest of the body.61 This assumption may hold in the able-bodied population; however, most persons with paraplegia depend on wheelchairs for ambulation. This results in loss of FFM in lower extremities caused by disuse atrophy,9, 62 with relative sparing in the upper extremities.11 Not surprisingly, resistance of the arm was lower (indicating greater TBW and therefore FFM in this body segment), reactance was higher (indicating greater BCM), and the correlation between (lengtharm)2/Rarm and TBW was weaker in the group with paraplegia. And as expected in this group, resistance of the leg was higher and reactance lower, indicating less TBW and BCM in the leg, respectively.

Limitations 

Our study contributes knowledge to some gaps in the literature, but 1 limitation bears mentioning. We used the Pace and Rathbun40 lean tissue hydration constant of 73.2% to determine FFM and BCM. This may or may not be appropriate in paraplegia. To determine a hydration constant specific to persons with paraplegia, we would have had to measure TBW as well as FFM by a second independent method, which was beyond the scope of our study. However, there is no reason to believe that the hydration constant of lean tissue in this population should differ from that of an able-bodied population. There is evidence to indicate that hydration of the fat-free body ranges from 72.7% for groups with low neurologic level of paraplegia to 76.0% to groups with low neurologic level of quadriplegia.8 Wang et al63 reported that the hydration constant is quite robust, even in situations of varying adiposity, and concluded that the change in TBW:FFM may be too small to identify using available in vivo methods. Nonetheless, determination of lean tissue hydration in men and women with complete versus incomplete and high versus low SCI is needed.

Back to Article Outline

Conclusions 

The widely used obesity BMI cutoff of ≥30kg/m2 has excellent specificity but poor sensitivity in distinguishing obese from nonobese subjects with paraplegia. This suggests that individuals with paraplegia and a BMI ≥30kg/m2 are likely truly obese, whereas no conclusions about body fatness can be made for those with a BMI less than 30kg/m2. We conclude that body composition should be measured in this population, and we present a validated prediction equation to do so using single-frequency BIA. We also present 2 additional equations to predict TBW (and subsequently FFM and FM), both of which performed well, but which need to be prospectively tested and cross-validated in an independent sample of men and women with chronic paraplegia.

Suppliers

Back to Article Outline

References 

  1. Duckworth W, Solomon S, Jallepalli P, Heckemeyer C, Finnern J, Powers A. Glucose intolerance due to insulin resistance in patients with spinal cord injuries. Diabetes. 1980;29:906–910
  2. Karlsson A, Attvall S, Jansson P, Sullivan L, Lonnroth P. Influence of the sympathetic nervous system on insulin sensitivity and adipose tissue metabolism (a study in spinal cord-injured subjects). Metabolism. 1995;44:52–58
  3. Bauman W, Spungen A, Raza M, et al.  Coronary artery disease (metabolic risk factors and latent disease in individuals with paraplegia). Mt Sinai J Med. 1992;59:163–168
  4. Krum H, Howes L, Brown D, et al.  Risk factors for cardiovascular disease in chronic spinal cord injury patients. Paraplegia. 1992;30:381–388
  5. Anson C, Shepherd C. Incidence of secondary complications in spinal cord injury. Int J Rehabil Res. 1996;19:55–66
  6. Boninger M, Cooper R, Baldwin M, Shimada S, Koontz A. Wheelchair pushrim kinetics (body weight and median nerve function). Arch Phys Med Rehabil. 1999;80:910–915
  7. Cardus D, McTaggart B. Total body water and its distribution in men with spinal cord injury. Arch Phys Med Rehabil. 1984;65:509–512
  8. Rasmann-Nuhlicek D, Spurr G, Barboriak J, Rooney C, El Ghatit A, Bongard R. Body composition of patients with spinal cord injury. Eur J Clin Nutr. 1988;42:765–773
  9. Jones L, Goulding A, Gerrard D. DEXA (a practical and accurate tool to demonstrate total and regional bone loss, lean tissue loss and fat mass gain in paraplegia). Spinal Cord. 1998;36:637–640
  10. Shizgal H, Roza A, Leduc B, Drouin G, Villemure JG, Yaffe C. Body composition in quadriplegic patients. JPEN J Parenter Enteral Nutr. 1986;10:364–368
  11. Wilmet E, Ismail A, Heilporn A, Welraeds D, Bergmann P. Longitudinal study of the bone mineral content and of soft tissue composition after spinal cord section. Paraplegia. 1995;33:674–677
  12. Cardus D, McTaggart B. Body composition in spinal cord injury. Arch Phys Med Rehabil. 1985;66:257–259
  13. Monreal M, Martorell A, Callejas JM, et al.  Venographic assessment of deep vein thrombosis and risk of developing post-thrombotic syndrome (a prospective study). J Intern Med. 1993;233:233–238
  14. Seem E, Stranden E, Stiris M. Computed tomography in deep venous thrombosis with limb oedema. Acta Radiol Diagn (Stockh). 1985;26:727–730
  15. Mammen EF. Pathogenesis of venous thrombosis. Chest. 1992;102(6 Suppl):640S–644S
  16. Bedogni G, Severi S, Manzieri AM, Trunfio O, Poli M, Battistini N. Use of bioelectric impedance analysis (BIA) in children with alterations of body water distribution. Appl Radiat Isot. 1998;49:619–620
  17. Nau K, Dick A, Peters K, Schloerb P. Relative validity of clinical techniques for measuring the body composition of persons with amyotrophic lateral sclerosis. J Neurol Sci. 1997;152(Suppl 1):S36–S42
  18. Paton N, Elia M, Jennings G, Ward L, Griffin G. Bioelectrical impedance analysis in human immunodeficiency virus-infected patients (comparison of single frequency with multifrequency, spectroscopy, and other novel approaches). Nutrition. 1998;14:658–666
  19. Schols A, Wouters E, Soeters P, Westerterp K. Body composition by bioelectrical-impedance analysis compared with deuterium dilution and skinfold anthropometry in patients with chronic obstructive pulmonary disease. Am J Clin Nutr. 1991;53:421–424
  20. Brummer R, Bengtsson B, Bosaeus I. Validation of body composition determination by bioelectrical impedance analysis in acromegaly. Eur J Clin Nutr. 1992;46:47–52
  21. Cardus D, McTaggart W. Electric impedance measurements in quadriplegia. Arch Phys Med Rehabil. 1988;69:186–187
  22. Spungen A, Bauman W, Wang J, Pierson R. Measurement of body fat in individuals with tetraplegia (a comparison of eight clinical methods). Paraplegia. 1995;33:402–408
  23. World Health Organization . Obesity (preventing and managing the global epidemic. Report of WHO Consultation). Geneva: WHO; 2000;
  24. Bray G. Health hazards of obesity. Endocrinol Metab Clin North Am. 1996;25:907–919
  25. George C, Wells C, Dugan N, Hardison R. Hydrostatic weights of patients with spinal injury (Reliability of measurements in standard sit-in and Hubbard tanks). Phys Ther. 1987;67:921–925
  26. Monroe M, Tataranni P, Pratley R, Manore M, Skinner J, Ravussin E. Lower daily energy expenditure as measured by a respiratory chamber in subjects with spinal cord injury compared with control subjects. Am J Clin Nutr. 1998;68:1223–1227
  27. Sedlock D, Laventure S. Body composition and resting energy expenditure in long term spinal cord injury. Paraplegia. 1990;28:448–454
  28. Green D, Twardowski P, Wei R, Rademaker A. Fatal pulmonary embolism in spinal cord injury. Chest. 1994;105:853–855
  29. Maki K, Briones E, Langbein W, et al.  Associations between serum lipids and indicators of adiposity in men with spinal cord injury. Paraplegia. 1995;33:102–109
  30. Burns S, Little J, Hussey J, Lyman P, Lakshminarayanan S. Sleep apnea syndrome in chronic spinal cord injury (associated factors and treatment). Arch Phys Med Rehabil. 2000;81:1334–1339
  31. Huang TS, Wang YH, Chen SY. The relation of serum leptin to body mass index and to serum cortisol in men with spinal cord injury. Arch Phys Med Rehabil. 2000;81:1582–1586
  32. Spungen A, Wang J, Pierson R, Bauman W. Soft tissue body composition differences in monozygotic twins discordant for spinal cord injury. J Appl Physiol. 2000;88:1310–1315
  33. Frisancho A. In: Methods and materials. Anthropometric standards for the assessment of growth and nutritional status. Ann Arbor: Univ Michigan Pr; 1990;p. 9–30
  34. Lukaski H, Johnson P, Bolunchuk W, Lykken G. Assessment of fat-free mass using bioelectrical impedance measurements of the human body. Am J Clin Nutr. 1985;41:810–817
  35. Baumgartner R, Chumlea W, Roche A. Estimation of body composition from bioelectric impedance of body segments. Am J Clin Nutr. 1989;50:221–226
  36. Buchholz A, Rafii M, Pencharz P. Is resting metabolic rate different between men and women?. Br J Nutr. 2001;86:641–646
  37. Schoeller DA, van Santen E, Peterson D, Dietz W, Jaspen J, Klein P. Total body water measurements in humans with 18O and 2H labeled water. Am J Clin Nutr. 1980;33:2686–2693
  38. Vaisman N, Pencharz PB, Koren G, Johnson J. Comparison of oral and intravenous administration of sodium bromide for extracellular water measurements. Am J Clin Nutr. 1987;46:1–4
  39. Scrimgeour C, Rollo M, Mudambo S, Handley L, Prosser S. A simplified method for deuterium/hydrogen isotope ratio measurements on water samples of biological origin. Biol Mass Spec. 1993;22:383–387
  40. Pace N, Rathbun E. Studies on body composition. III. The body water and chemically combined nitrogen content in relation to fat content. J Biol Chem. 1945;158:685–691
  41. Jervis R, Hancock R, Hill D, Isles K. Biomedical and health studies with the new Canadian Slowpoke Reactor. J Radioanal Chem. 1977;37:463–471
  42. Lohman T, Going S. Assessment of body composition and energy balance. In:  Lamb D,  Murray R editor. Perspectives in exercise science and sports medicine. Carmel: Cooper Publishing Group; 1998;p. 61–105
  43. Bland J, Altman D. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;1:307–310
  44. Yao M, Roberts S, Ma G, Pan H, McCrory M. Field methods for body composition assessment are valid in healthy Chinese adults. J Nutr. 2002;132:310–317
  45. Piers L, Soares M, Frandsen S, O’Dea K. Indirect estimates of body composition are useful for groups but unreliable in individuals. Int J Obes Relat Metab Disord. 2000;24:1145–1152
  46. Israel R, Pories W, O’Brien K, McCammon M. Sensitivity and specificity of current methods for classifying morbid obesity. Diabetes Res Clin Pract. 1990;10(Suppl 1):S145–S147
  47. Bedogni G, Merlini L, Ballestrazzi A, Severi S, Battistini N. Multifrequency bioelectric impedance measurements for predicting body water compartments in duchenne muscular dystrophy. Neuromuscul Disord. 1996;6:55–60
  48. Deurenberg P, Wolde-Gebriel Z, Schouten F. Validity of predicted total body water and extracellular water using multifrequency bioelectrical impedance in an Ethiopian population. Ann Nutr Metab. 1995;39:234–241
  49. Segal K, van Loan M, Fitzgerald P, Hodgdon J, van Itallie T. Lean body mass estimation by bioelectrical impedance analysis (a four-site cross-validation study). Am J Clin Nutr. 1988;47:7–14
  50. NIH Consensus statement. Bioelectrical impedance analysis in body composition measurement. National Institutes of Health Technology Assessment Conference Statement. December 12–14, 1994. Nutrition. 1996;12:749–759
  51. Schoeller D. Bioelectrical impedance analysis (what does it measure?). Ann N Y Acad Sci. 2000;904:159–162
  52. Cha K, Chertow G, Gonzalez J, Lazarus J, Wilmore D. Multifrequency bioelectrical impedance estimates the distribution of body water. J Appl Physiol. 1995;79:1316–1319
  53. Cornish B, Ward L, Thomas B, Jebb S, 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;50:159–164
  54. Simpson J, Lobo D, Anderson J, et al.  Body water compartment measurements (a comparison of bioelectrical impedance analysis with tritium and sodium bromide dilution techniques). Clin Nutr. 2001;20:339–343
  55. Segal K, Burastero S, Chun A, Coronel P, Pierson R, Wang J. Estimation of extracellular and total body water by multiple-frequency bioelectrical-impedance measurement. Am J Clin Nutr. 1991;54:26–29
  56. Sergi G, Bussolotto M, Perini P, et al.  Accuracy of bioelectrical impedance analysis in estimation of extracellular space in healthy subjects and in fluid retention states. Nutr Metab. 1994;38:158–165
  57. Van Loan M, Mayclin P. Use of multi-frequency bioelectrical impedance analysis for the estimation of extracellular fluid. Eur J Clin Nutr. 1992;46:117–124
  58. van Marken-Lichtenbelt W, Westerterp K, Wouters L, Luijendijk S. Validation of bioelectrical-impedance measurements as a method to estimate body-water compartments. Am J Clin Nutr. 1994;60:159–166
  59. Hannan W, Cowen S, Plester C, Fearon K, DeBeau A. Comparison of bio-impedance spectroscopy and multi-frequency bio-impedance analysis for the assessment of extracellular and total body water in surgical patients. Clin Sci. 1995;89:651–658
  60. Kushner R. Bioelectrical impedance analysis (a review of principles and applications). J Am Coll Nutr. 1992;11:199–209
  61. Stewart S, Bramley P, Heighton R, et al.  Estimation of body composition from bioelectrical impedance of body segments (comparison with dual-energy X-ray absorptiometry). Br J Nutr. 1993;69:645–655
  62. Baldi J, Jackson R, Moraille R, Mysiw W. Muscle atrophy is prevented in patients with acute spinal cord injury using functional electrical stimulation. Spinal Cord. 1998;36:463–469
  63. Wang Z, Deurenberg P, Wang W, Pietrobelli A, Baumgartner R, Heymsfield S. Hydration of fat-free body mass (review and critique of a classic body-composition constant). Am J Clin Nutr. 1999;69:833–841
  • 1 No commercial party having a direct financial interest in the results of the research supporting this article has or will confer a benefit upon the author(s) or upon any organization with which the author(s) is/are associated.
  • a Detecto model; Cardinal Scale Manufacturing Co, PO Box 151, 203 E Daugherty, Webb City, MO 64870.
  • b Model 6006; Scale-Tronix, 200 E Post Rd, White Plains, NY 10601.
  • c Harpenden, Holtain Ltd, Crymych, SA41 3UF, UK.
  • d Model 101; RJL Systems, 33955 Harper Ave, Clinton Township, MI 48035.
  • e Xitron Technologies Corp, 10225 Barnes Canyon Rd, Ste A102, San Diego, CA 92121.
  • f C/D/N Isotopes, 88 Leacock St, Pointe-Claire, QC H9R 1H1, Canada.
  • g Fisher Scientific Ltd, 112 Colonnade Rd, Nepean, ON K2E 7L6, Canada.
  • h J6B centrifuge; Beckman Coulter In, PO Box 3100, 4300 N Harbor Blvd, Fullerton, CA 92834-3100.
  • i Europa CF-IRMS 20/20; PDZ Europa Ltd, Hill St, Cheshire, CW11 3JE, UK.
  • j Version 8.1; SAS Institute Inc, 100 SAS Campus Dr, Cary, NC, 27513-2414.

☆ Supported in part by the Ontario Neurotrauma Foundation (grant no. ONBO-00026).

PII: S0003-9993(02)04950-X

doi:10.1016/S0003-9993(02)04950-X

Archives of Physical Medicine and Rehabilitation
Volume 84, Issue 6 , Pages 854-861, June 2003

Article Tools

* Image Usage & Resolution