Numerous researchers have examined the reliability and validity of bioelectrical impedance (BI) (3-14,17,20,21,23,24,26,28). A problematic issue with the validity of BI is that resistance measurements appear to be confounded by other variables, such as the hydration state of the subject or exercise (7,8,11). Researchers have shown that as individuals become dehydrated, bioelectrical impedance (BI) measurements may underestimate an individual's true body fat percentage (7,11). This is especially important as it relates to the measurement of athletes. For example, researchers have examined the use of BI in determining minimum weights for wrestlers during competition (17). Because wrestlers often use exercise or other methods of dehydration to reduce weight for competition, these alterations in hydration state could preclude accurate body composition measurements using BI. In addition, in situations where athletes are regularly measured, daily alterations in hydration (from sweating and fluid consumption) could affect the validity of percent fat measures.
The extent to which changes in hydration effect BI measures in athletes has not been adequately quantified. In addition, no studies have systematically investigated how the ionic make-up of the rehydration solution might effect BI results. For example, would the rehydration of an individual using an ionic rehydration solution alter the accuracy and consistency of BI testing to a greater level than water alone? Therefore, the purpose of this investigation was to determine how hypohydration and rehydration with either a nonelectrolyte solution (deionized water) or an electrolyte solution (Gatorade, Chicago, IL), and superhydration using both solutions, affected the validity of BI measurements in comparison to hydrostatic weighing.
Fifteen subjects (11 male and 4 female) aged 19 to 56 yr volunteered for the study. All subjects were engaged in a running program and performing a long run of 50 min or greater at least once every 2 wk. Since the main purpose of this study was to look at how hydration state changes affected BI and hydrostatic weighing (HW) body composition values, not how gender or age played a role, male and female subjects were collapsed into one group serving as their own controls in this repeated design study. Based on previous work in our lab, a minimum of eight subjects were required for adequate statistical power to determine any between-group differences (power level = 0.90 when P = 0.05). No subjects were allowed to participate who had any cardiovascular (e.g., hypertension) or metabolic (e.g., diabetes) related disorders. All procedures were approved by the East Tennessee State University Institutional Review Board; written informed consent was obtained from all subjects.
Graded treadmill test. Subjects performed a running protocol to determine maximal aerobic capacity, in which speed and/or grade were increased on a Quinton 55 treadmill (Quinton Instrument Co., Seattle, WA) every 1-2 min following a 3-min warm-up at 3.0 mph and 0% grade. Subjects breathed through a two-way mouthpiece interfaced to a SensorMedics 2900 (Anaheim, CA) metabolic cart to analyze expired gases. Subjects also wore a nose piece to ensure that all expired gases were collected by the analyzer. Speed and/or grade were increased until each subject reached volitional fatigue. Exercising heart rates during testing were measured using a Polar Vantage XL (Port Washington, NY) heart rate monitor. Maximal aerobic capacity, as measured by V˙O2max, was determined by the average of the three highest 20-s V˙O2 intervals during the test when the respiratory exchange ratio exceeded 1.1 and the maximum heart rate was ± 10 beats · min−1 of each subject's age predicted maximum.
Hydrostatic weighing and residual volume. Hydrostatic weighing was performed on each subject as described by Benke and Wilmore (1) to determine body density. Subjects were seated in chairs suspended from a spring-loaded, 9-kg scale in a stainless steel tank. Subjects were instructed to submerge while expiring maximally and remain as motionless as possible at the point of maximal expiration for ∼3 s while underwater weight was recorded. A minimum of seven trials were performed until a clear plateau in underwater weight was observed. The estimated underwater weight was determined by the mean of the highest three recorded values within 0.025 kg. Water temperature and tare weight were recorded before each trial. Body density was determined using the equation of Benke and Wilmore (1).
Residual lung volume was estimated using the modified O2-dilution method of Wilmore et al. (27). Two trials were performed out of water while the subjects assumed a sitting position that replicated the body position in the tank during underwater weighing. The average of the two trials was used to determine percentage of nitrogen. However, if the values were not within 1%, a third trial was performed. This procedure has been shown by Wilmore et al. (28) to be both reliable (r = 0.99) and valid (r = 0.92) for determining residual lung volume. Percent fat was calculated using the Siri (25) equation for all subjects.
Height and weight. Height was measured in centimeters and rounded to the nearest half centimeter using a standard height measure (Detecto Inc., Brooklyn, NY). Nude weight was determined in kilograms using a Portatronic digital scale (General Inc., Cape Coral, FL). Measurements were rounded to the nearest 0.1 kg.
Bioelectrical impedance. BI was performed using the Valhalla Scientific 1990B (San Diego, CA) analyzer. Subjects lay supine with their arms at their sides, away from their bodies. They were instructed to remove their shoes, socks, and jewelry. Aluminum foil spot electrodes (Unitrace, United Medical Supply Corp., Littleton, CO) were placed on the hands and feet of the subject's dominant side. Hand electrodes were positioned at the line bisecting the styloid processes of the ulna and radius and on the posterior surface of the hand at the distal metacarpal. Foot electrodes were placed in a line bisecting the medial and lateral malleoli and on the dorsal surface of the foot at the distal metatarsal. Following electrode placement, the subject's height, weight, age, sex, and activity level (for females only) were entered into the computer. Resistance, reactance, impedance, and phase were calculated by the analyzer. These variables were used to determine fat-free mass, fat mass, and percent fat using the manufacturers equations (R2 = 95.6; SEE 2.6).
Exercise bout. The exercise bout consisted of a run on a treadmill in an environmental chamber with the temperature set at 85°F and 50% humidity. Subjects were instructed to run at a self-selected pace that corresponded to 50-60% of their V˙O2max (as determined by the graded treadmill test). However, subjects were permitted to run faster than this pace if they desired. Subjects had their nude weight measured pre-exercise, 30 min after beginning exercise, and every 10 min thereafter. The exercise bout was ended when the subject had lost approximately 3% of his/her nude body weight because of sweating.
Rehydration. Subjects were rehydrated using either deionized water or Gatorade. They were instructed to drink the same weight of fluid as they lost through exercise. When they were using Gatorade, 70 g of Gatorade power was mixed with each liter of water consumed. As a result of manipulating body weight 3% of a person's normally hydrated state, fluid ingestion during rehydration and suprahydration interventions resulted in mean fluid intakes of 1.89 liters per person.
Assessment protocol. Subjects in the study completed one pretreatment trial and four treatment trials. During all trials subjects reported to the lab at least 3 h after a light meal. Each subject consumed approximately 500 mL of water 2 h pretrial to standardize hydration levels. Subjects were encouraged to empty their bowels before the trial began since they were not permitted to do so once the trial was underway. Subjects completed all four treatment trials within a 15-d period.
Experimental design. Each subject performed a pretreatment trial consisting of a graded exercise test and a hydrostatic weighing measurement. This trial was performed to determine running speed during the trials and to familiarize subjects with the hydrostatic weighing procedure. In addition, subjects performed four randomly assigned treatment trials, which consisted of a battery of tests (including weight, hydrostatic weighing with RV measurement, and BI measures) before and after hydration manipulations. During two trials, subjects had their hydration states manipulated twice: once by an exercise bout (hypohydration) and once with an ingestion of fluid (rehydration) (see Fig. 1). These two trials differed only in the fluid ingested (once with deionized water, once with Gatorade). For the remaining two trials, subjects had their hydration states manipulated only with ingestion of water or Gatorade (superhydration).
Statistical analyses. For statistical purposes, the results of this study were divided into two substudies: (a) a superhydration study and (b) a hypohydration/rehydration study. The superhydration study involved comparisons between BI and HW at normal versus superhydration. Regression analyses were used to determine test-retest reliability and the relationship between BI and HW. An ANOVA model was used to determine differences between percent fat measures for both techniques, and differences between superhydrating solutions (deionized water and Gatorade). A 2 × 2 repeated measures ANOVA model was used to determine differences between measurement techniques across the hydration states.
The hypohydration/rehydration study involved comparisons between BI and HW over three hydration states: normal, hypohydrated, and rehydrated. Regression analyses were used to determine test-retest reliability and the relationship between HW and BI. A 3 × 2 repeated measures ANOVA model was used to determine differences between BI and HW % fat across hydration states. In addition, an ANOVA model was used to assess changes in % fat, fat-free weight (FFW), and fat-weight (FW) for each technique across hydration states. Multiple group orthogonal comparisons were used to determine where any differences occurred. An ANOVA model was also used to determine differences in rehydrating solutions.
Superhydration study. The physical demands of the exercise bout in the hypohydration/rehydration study required that subjects were generally young, fit, athletic individuals. This is demonstrated by the subjects' descriptive statistics (Table 1). These mean values were calculated at normal hydration and were based on two trials for each subject: the non-exercise/water trial and the non-exercise/Gatorade trial combined.
One initial interest of the study was to determine what effect electrolyte content of the hydrating solution would have on BI's validity in determining relative body fat. There were no significant differences (P = 0.42) in BI percent fat (15.4 ± 5.4%; 16.0 ± 5.5%) or HW percent fat (13.5 ± 6.7%; 13.4 ± 6.7%) between the water and Gatorade trials, respectively, following superhydration. Furthermore; the increases in percent fat from normal to superhydration were very similar for BI (2.1%; 2.2%) and HW (0.6%; 0.6%) between water and Gatorade trials. Consequently, BI and HW measurements for both hydration states (normal and superhydrated) were calculated as mean values of the water and Gatorade trials.
Hydrostatic weighing was shown to be extremely reliable across the two trials at normal hydration (r = 0.99; SEE = 0.5%) (P < 0.0001). BI was also reliable (r = 0.93), but had greater error (SEE = 1.4%) than HW (P < 0.0001). In addition, the mean values for HW (12.9 ± 7.0%; 12.8 ± 6.9%) and BI (13.9 ± 5.0%; 13.2 ± 5.3%) did not significantly differ between the two trials at normal hydration (P = 0.20).
Bioelectrical impedance was compared with hydrostatic weighing in both normal and superhydrated states. Regression analyses, performed for percent fat values for BI against HW at normal hydration, showed a moderate relationship (r = 0.82; SEE = 3.1%) (P < 0.0002) between the measures. The mean values for BI (13.6 ± 5.1%) were not significantly different than those for HW (12.8 ± 6.9%) (P = 0.49).
BI and HW were more strongly correlated (r = 0.89; SEE = 2.7%) in the superhydrated state (P ≤ 0.0001). However, there was a significant difference (P ≤ 0.017) between the mean values for BI percent fat (15.7 ± 5.5%) versus HW percent fat (13.4 ± 6.9%). This resulted in a 2.3% absolute overprediction of percent fat by BI in comparison to HW in the superhydrated state.
HW percent fat did not vary significantly from normal hydration (12.8 ± 6.9%) to superhydration (13.4 ± 6.9%) (P ≤ 0.10). By contrast, there was a significant increase in BI percent fat values following superhydration (15.4 ± 5.6%) versus normal hydration (13.2 ± 5.3%) (P ≤ 0.0001) (Fig. 2).
At normal hydration, BI determined FFW was significantly related (P ≤ 0.0001) to HW FFW (r = 0.97; SEE = 5.0 kg). BI FFW (54.4 ± 6.5 kg) was not significantly different (P ≤ 0.49) than HW FFW (55.0 ± 8.1 kg). In addition, BI FW at normal hydration was also significantly related (P ≤ 0.0002) to HW FW (r = 0.83; SEE = 1.7 kg). Means for BI FW (8.6 ± 3.8 kg) were not significantly different (P ≤ 0.39) from HW FW (8.1 ± 4.5 kg) at normal hydration.
Although BI FFW and FW were not significantly different than HW FFW and FW at normal hydration, the superhydration intervention altered the normal hydration similarities. Superhydrated BI FFW was still strongly correlated (P < 0.0002) to HW FFW (r = 0.97; SEE = 4.2 kg). However, the mean values for superhydrated BI FFW (54.5 ± 7.0 kg) were significantly less than HW FFW (56.2 ± 8.2 kg) (P ≤ 0.009). Accordingly, superhydrated BI FW had a significant relationship (p ≤ 0.0001) with HW FW (r = 0.88; SEE = 1.7 kg), but mean values for superhydrated BI FW (10.2 ± 3.9 kg) were significant higher (P ≤ 0.02) than mean values for superhydrated HW FW (8.7 ± 4.7 kg).
The mean fluid weight gain during the trials was 1.8 ± 0.2 kg. A vast majority of this fluid weight was interpreted as fat weight by bioelectrical impedance. Superhydrated fat weight increased 1.53 kg for BI versus 0.63 kg for HW following fluid intake. Therefore, 85% of the superhydrating fluid was interpreted as fat by BI versus 35% by HW (Fig. 3).
The slopes and intercepts from regression analyses of BI and HW normal and superhydrated states were compared. The BI intercept (1.59) was approximately four times greater than the HW intercept (0.372) for the regression analysis. In contrast, the slopes were similar for both BI (0.994) and HW (1.0). While both HW and BI had elevated % fat predictions during superhydration, only the BI measurement was significantly greater during superhydration.
Hypohydration/rehydration study. Similar to the superhydration study, the hydrating solution (water and Gatorade, respectively) did not have a significant effect (P ≤ 0.33) on BI percent fat between normal (13.7 ± 5.4%; 14.3 ± 5.2%), hypohydrated (12.2 ± 6.0%; 12.3 ± 4.8%), and rehydrated (15.1 ± 6.6%; 15.8 ± 5.4%) hydration states. In addition, there were not any significant differences (P ≤ 0.48) in HW percent fat values using water or Gatorade, respectively, at normal (13.3 ± 7.1%; 13.1 ± 7.3%), hypohydrated (13.4 ± 7.5%; 13.5 ± 7.5%), or rehydrated (13.5 ± 7.3%; 13.6 ± 7.4%) hydration states. Accordingly, BI and HW measurements for all hydration states (normal, hypohydrated, and rehydrated) were calculated as mean values of the water and Gatorade trials.
HW was highly reliable between trials at normal hydration (r = 0.98, SEE = 0.8%) (P ≤ 0.0001). BI was also demonstrated to be reliable across trials (r = 0.96; SEE = 1.1%) (P ≤ 0.0001). In addition, the mean values for HW (13.3 + 7.1%; 13.1 + 7.3%) and BI (13.7 ± 5.4%; 14.3 ± 5.2%) did not significantly differ (P ≤ 0.47) between the two trials at normal hydration.
Regression analyses were performed to determine the relationships across hydration states independently for each percent fat assessment technique (BI or HW). The results showed a significant relationship (P ≤ 0.0001) for the HW percent fat values between normal hydration and hypohydrated states (r = 0.997; SEE = 0.3%), as well as between normal and rehydrated states (r = 0.997; SEE = 0.3%) (P ≤ 0.0001).
BI percent fat measures were significantly related (P ≤ 0.0001) between normal hydration and hypohydration (r = 0.94; SEE = 1.2%), and between normal hydration and rehydration states (r = 0.92; SEE = 1.6%) (P ≤ 0.0001). In addition, at normal hydration, BI was significantly related (P ≤ 0.001) to HW (r = 0.75; SEE = 3.7%). BI was also significantly related to HW at both hypohydration (r = 0.71; SEE = 3.7%) (P ≤ 0.003) and rehydration (r = 0.74, SEE = 3.8%) (P ≤ 0.002).
HW percent fat values did not significantly differ (P ≤ 0.12) between normal (13.2 ± 7.2%), hypohydrated (13.5 ± 7.5%), or rehydrated states (13.6 ± 7.3%). As a result, HW percent fat values remained stable across all hydration levels. By contrast, the mean values for BI percent fat significantly decreased from normal (14.0 ± 5.2%) to hypohydration (12.3 ± 5.3%) (P ≤ 0.001), and significantly increased between hypohydrated and rehydrated (15.5 ± 5.8%) hydration states (P ≤ 0.0001). In addition, BI percent fat significantly increased following rehydration compared with that during normal hydration (P ≤ 0.01) (Fig. 4).
Mean values for HW FW were significantly different between hydration states (P ≤ 0.006). This was expected because of the changes in total body weight. Multiple comparison orthogonal post-hoc analyses showed a significant increase in HW FW between normal (8.4 ± 4.8 kg) and rehydrated (8.6 ± 4.9 kg) hydration states (P ≤ 0.02), and between hypohydrated (8.3 ± 4.8 kg) and rehydrated states (P ≤ 0.02). Mean values for HW FFW were also significantly different between hydration states (P ≤ 0.0001). HW FFWs significantly decreased between normal (55.0 ± 8.1 kg) and hypohydrated (53.3 ± 8.0 kg) hydration states (P ≤ 0.0001), and significantly increased between hypohydrated and rehydrated (54.8 ± 8.0 kg) states (P ≤ 0.0001).
BI FWs were significantly different across hydration states (P ≤ 0.0001). Multiple comparison post-hoc analyses showed that BI FWs significantly decreased between normal (8.9 ± 4.0 kg) and hypohydrated (7.5 ± 3.7 kg) hydration states (P ≤ 0.0001). Conversely, BI FW significantly increased from normal to rehydration (9.8 ± 4.4 kg) (P ≤ 0.006), and from hypohydration to rehydration (P ≤ 0.0001). Interestingly, BI FFWs were not significantly different (P ≤ 0.80) between normal (53.9 ± 6.9 kg), hypohydrated (53.9 ± 6.3 kg), and rehydrated (53.6 ± 6.3 kg) hydration states.
The mean weight loss from hypohydration was 1.7 ± 0.3 kg. The mean weight gain from rehydration was 1.8 ± 0.3 kg. Therefore, 82% of the water (sweat) lost during hypohydration was perceived as FW loss by BI versus 6% by HW. Furthermore, 128% of the fluid weight gained through rehydration was interpreted as FW by BI compared with 17% by HW (Fig. 5).
At normal hydration BI (14.0 ± 5.2%) overpredicted percent fat compared with HW (13.2 ± 7.2%) (P ≤ 0.05). Following hypohydration, BI (12.3 ± 5.3%) underpredicted percent fat versus HW (13.5 ± 7.3%) (P ≤ 0.003). Rehydration reversed this effect, causing BI (15.5 ± 5.8%) to overestimate again % fat compared with HW (13.6 ± 7.3%) (P ≤ 0.0001).
The results of this study confirmed the variability of BI measures during periods of altered hydration seen by other researchers (7,11). When data were compared at normal hydration, the mean % fats for BI (13.6 ± 5.1%, 14.0 ± 5.2%) were within 1% of the mean values for HW (12.8 ± 6.9%, 13.2 ± 7.2%). However, the SEs associated with BI (3.1-3.7%) precluded its ability to predict all subjects accurately. In fact, at normal hydration BI predicted only 63% of subjects within 4% of HW values. Previous researchers have shown BI predictability to range from 48% (26) to 72% (28) using a 4% mean difference criteria. Percent fat values obtained by HW remained stable across all hydration states, not significantly changing between normal hydration, hypohydration, and rehydration (P ≤ 0.12), or from normal to superhydration (P ≤ 0.10). By contrast, BI percent fat values significantly decreased 1.7% from normal to hypohydration, increased 3.2% from hypohydration to rehydration, and increased 2.2% from normal to superhydration. These findings confirmed the results of Gilliam (7) who reported increases in BI percent fat values following consumption of water (27.3 + 9.7%) compared with normal hydration (26.7 + 9.6%) in college females. Gilliam also found a decrease in BI percent fat values following dehydration through caffeine ingestion (25.7 + 9.6%). In addition, Khaled et al. (11) reported that dehydration resulted in decreased resistance values (and subsequently decreased percent fat values) while overhydration resulted in increased resistance values (accompanied by increase percent fat values) using BI. Similar to Khaled et al. (11), the increases in percent fat with fluid intake in this study were accompanied by increased resistance values during both rehydration (531-553 Ω) and superhydration (532-538 Ω). However, the decreased percent fat following hypohydration occurred despite an increase in resistance (from 521 to 531 ohms). This was apparently a result of the significant contribution of body weight in the BI regression equation, which appears to interpret fluid loss/gain as fat weight loss/gain. Other investigators have reported similar results with respect to changes in resistance with fluid ingestion. Gomez et al. (9) reported a 6.4 to 10.4 Ω increase in resistance following water intake measured 4-90 min after ingestion. Similarly, Rising et al. (19) reported R-values that were elevated 10 Ω following 700 mL of fluid ingestion in Pima Indians. Interestingly, however, the fluid ingested in this study did not translate into a detectable change in percent body fat. This is probably because the relative body weight change with fluid intake in the study of Rising et al. was much smaller than the present study (because of ingestion of less fluid (0.7 vs 1.89 L) and larger body weights of subjects (91.7 vs 63.0 kg)). This observation provides more evidence for the hypothesis that significant fluid weight changes are largely interpreted as fat weight changes by BI regression equations. In contrast with the aforementioned studies and the present study is data from Deurenberg et al. (3). These researchers showed that mean body impedance decreased 4 Ω 20 min after ingesting beef tea. In addition, they reported that impedance decreased 9 Ω following strenuous exercise for 90 min. This discrepancy from the present study can potentially be explained by one of two differences in methodology. First, resistance has been shown by Caton et al. (2) to decrease when skin is warmed. If the BI measurements in the study of Deurenberg et al. (3) were performed immediately after exercise, it is possible that skin temperature alone reduced impedance. In the present study this factor was partially controlled by taking BI measurements 40-45 min after exercise or fluid ingestion (10-15 min allotted), and this measurement always followed the HW measurement, in which subjects spent approximately 15 min in a tank of water at a constant temperature. In addition, fluid was consumed at room temperature to reduce the effect of cooling by the liquid. The second methodological difference between the studies is that the subjects in the Deurenberg et al. study (3) were permitted to drink ad libitum during exercise and no changes in body weight were measured. Therefore, it is difficult to determine what independent effect exercise had on resistance under such conditions.
A 2-3% change in body fat percentage may not be of great practical importance for normal populations. However, in this athletic population, 2-3% fat represented 16-23% of total body fat stores. This is of practical importance because athletes often manipulate their eating habits based on their body composition. For example, if a male distance runner's percent body fat was estimated to be 11% instead of 8% during the competitive season, he would possibly restrict his diet unnecessarily to reduce his percent fat. In extreme cases, this form of inaccurate feedback could start or reinforce disordered eating patterns.
A number of researchers have discussed the advantages of applying population specific equations to BI data (4,10,13,20,23,24). The results of these studies suggests that population-specific equations may improve the ability of BI to predict percent fat and FFW. Manufacturer's equations were employed in our study to simulate the conditions that were most likely to be employed in an applied sports setting. However, for comparative purposes, we applied population-specific equations for athletic males (18) and athletic females (15) to our data. These equations showed the same trends as the manufacturers' equations across hydration states (a decrease in percent fat following hypohydration and an increase in percent fat following rehydration or superhydration). However, the magnitude of the changes in percent fat from normal (14.2 ± 5.2%) to hypohydration (12.3 ± 5.3%) to rehydration (15.5 ± 5.8%) and from normal (13.2 ± 5.3%) to superhydration (15.4 ± 5.6%) were smaller than those observed when manufacturers' equations were employed. This observation supports the use of population specific equations over manufacturers' equations in conditions of altered hydration.
BI-estimated FFWs remained very stable during the study, with no significant differences between any hydration states. This was a somewhat unexpected finding. Because there were significant changes in total body weights between hydration states, it was anticipated that a portion of weight loss or gain would come from both FFW and FW. This was demonstrated in HW, where there were proportionally equal differences between FFW and FW following each hydration change.
Changes in BI FW following each hydration change were much larger than expected in comparison with those in HW. A mean weight loss (through hypohydration) of 1.7 ± 0.3 kg resulted in a 1.4-kg decrease in BI-determined FW, compared with a 0.1 kg decrease in HW FW. A 1.8 ± 0.3 kg weight gain following rehydration showed a 2.3 kg increase in BI FW versus a 0.3 kg increase in FW by HW. Finally, a superhydration weight gain of 1.8 ± 0.2 kg resulted in a 1.6 kg increase in BI FW compared to 0.6 kg for HW.
A technique that was 100% accurate across hydration states would hypothetically show a change in FW and FFW that was proportional to the body fat percentage of the individual. Therefore, under optimal conditions BI FWs should have decreased 0.2 kg (to 8.7 kg) following hypohydration and increased 0.3 kg (to 9.0 kg) after rehydration (based on a mean weight of 63.4 kg and BI percent fat of 14.0% during the hypohydration/rehydration study). During the superhydration trial, BI FWs should have increased 0.2 kg (to 8.8 kg) based on the mean values for weight (63.0 kg) and BI percent fat (13.6). Based on these assumptions, BI FW was underestimated following hypohydration (by 1.2 kg), and overestimated following rehydration (by 0.8 kg) and following superhydration (by 1.4 kg). This suggests that any changes in weight because of hydration alterations were perceived as fat weight changes by BI.
The relatively large fluctuations in BI FW after treatments are of special interest as they relate to observations by Gilliam (7) that increases in water would theoretically be expected to reduce resistance values and therefore be perceived as additional lean tissue, while decreases in water would be perceived as additional fat. Similarly, Segal (22) suggested that decreases in body water from sweating would increase resistance (and percent fat) measured by BI. Gilliam, whose results paralleled the present study, concluded that the BI machine was responding to the changes in body weight rather than changes in water. While the present study appears to support this hypothesis, it is important to note that following a 1.8-kg rehydration, BI FW increased 2.3 kg. Therefore, the changes in hydration affected percent fat values beyond that which could be explained by changes in weight alone. Some of the differences between the magnitude of this study's results versus Gilliam's may be explained by observations from Matthews and Gilker (16). These investigators noted that the relationship between FFW and total body water (TBW), FFW = TBW/0.73, may be in error by several percent depending on hydration levels and distribution of water in the body. Matthews and Gilker (16) determined that when FW is assessed during a fluid imbalance where TBW no longer equals 0.73 * FFW (such as hypohydration, rehydration, or superhydration), a hydration error can significantly distort FW, especially in those subjects with low relative percentages of fat. BI measurements have been shown to be sensitive to changes in TBW (20,21). Therefore, it is likely that subjects in this study experienced greater distortions in BI determined FW (across hydration states) because their mean body fat percentages (12.8%) were much lower than those of Gilliam's subjects (21.7%).
The distribution of the ingested water within the body may have also been a factor in the increased BI FWs. The subjects in this study typically had BI measurements taken 40-45 min after ingestion of fluid. According to Matthews and Gilker (16), this was likely not enough time for the ingested fluid to equilibrate into intracellular and interstitial pools (approximately 2 h). The ingested fluid was probably still in the smaller extracellular, intravascular (plasma) compartment while BI measurements were taken (in both Gilliam's study (7) and the present study).
Blood volume measurements performed on the final five subjects during the hypohydration/rehydration study supports this hypothesis. Analyses of intravenous blood samples indicated that blood volumes increased 3.7 ± 1.8%, while plasma volumes increased 2.4 ± 4.3% from hypohydration to rehydration. These changes suggest that a majority of the ingested fluid was in the intravascular compartment. This hypothesis could also explain why the electrolyte content of the hydrating solution did not have any effect on the BI measurements. It would be valuable to investigate how BI FW values changed if measurements were taken two or more hours following ingestion of fluid (once equilibration had occurred). However, because a 3% change in body weight was established through pilot work in our lab for creating a measurable change, delaying the BI measurements beyond 45 min to allow proper water absorption to the tissue was not practical nor comfortable for the subjects. This obviously would change the subjects' body weights, which would have an effect on BI FW measures.
In summary, changes in hydration states caused large fluctuations in BI FWs, resulting in significant changes in BI-determined percent fat measurements. Therefore, to achieve reliable BI measurements, care must be taken to ensure standardization of hydration. This observation would appear to preclude the ability of BI to determine body composition changes in runners, wrestlers, or other athletes whose hydration states are likely to fluctuate across measurements. Future research should consider the level of equilibration of fluids in the body to determine the effect on BI measurements.
1. Benke, A. R. and J. H. Wilmore. Evaluation and regulation of body build and composition. Englewood Cliffs, NJ: Prentice-Hall, 1974.
2. Caton, J. R., P. A. Mole, W. C. Adams, and D. S. Henstis. Body composition analysis by bioelectrical impedance: effect of skin temperature. Med. Sci. Sports Exerc.
3. Deurenberg, P., J. A. Westrate, I. Paymans, and K. van der Kooy. Factors affecting bioelectrical impedance measurements in humans. Eur. J. Clin. Nutr.
4. Deurenburg, P., J. A. Weststrate, and K. van der Kooy. Body composition changes assessed by bioelectrical impedance measurements. Am. J. Clin. Nutr.
5. Eckerson, J. M., T. J. Housh, and G. O. Johnson. Validity of bioelectrical impedance equations for estimating fat-free weight in lean males. Med. Sci. Sports Exerc.
6. Fulcher, G. R., M. Farrer, M. Walker, D. Rodham, B. Clayton, and K. G. M. M. Alberti. A comparison of measurements of lean body mass derived from bioelectrical impedance, skin-fold thickness and total body potassium: a study in obese and non-obese normal subjects. Scand. J. Clin. Lab. Invest.
7. Gilliam, L. F. The effects of hydration levels upon bioelectrical impedance measures of body composition in college females. Unpublished thesis, Johnson City: East Tennessee State University, 1988.
8. Gleichauf, C. N., and D. A. Roe. The menstrual cycle's effect on the reliability of bioelectrical impedance measurements for assessing body composition. Am. J. Clin. Nutr.
9. Gomez, T., P. A. Mole, and A. Collins. Dilution of body fluid electrolytes affects bioelectrical impedance measurements, Sports Med., Training, Rehab.
10. Jackson, A. S., M. L. Pollock, J. E. Graves, and M. T. Mahar. Reliability and validity of bioelectrical impedance in determining body composition. J. Appl. Physiol.
11. Khaled, M. A., M. J. McCutcheon, S. Reddy, P. L. Pearman, G. R. Hunter, and R. L. Weinsier. Electrical impedance in assessing human body composition: the BIA method. Am. J. Clin. Nutr.
12. Lukaski, H. C., P. E. Johnson, W. W. Bolonchuk, and G. I. Lykken. Assessment of fat-free mass using bioelectrical impedance measurements of the human body. Am. J. Clin. Nutr.
13. Lukaski, H. C., W. W. Bolonchuk, C. B. Hall, and W. A. Siders. Validation of tetrapolar bioelectrical impedance method to assess human body composition. J. Appl. Physiol.
14. Lukaski, H. C. Methods for the assessment of human body composition: traditional and new. Am. J. Clin. Nutr.
15. Lukaski, H. C., and W. W. Bolonchuk. Theory and validation of the tetrapolar bioelectrical impedance method to assess human body composition. In: In Vivo Body Composition Studies.
K. J. Ellis, S. Yasamura, and W. D. Morgan (Eds.). London: The Institute of Physical Sciences in Medicine, 1987, pp. 410-414.
16. Matthews, D. W., and C. D. Gilker. Impact of 2
H and 18
O pool size determinations on the calculation of total energy expenditure. Obes. Res.
3(Suppl. 1):21-29, 1995.
17. Opplinger, R. A., D. H. Nielsen, and C. G. Vance. Wrestlers' minimum weight: anthropometry, bioimpedance, and hydrostatic weighing compared. Med. Sci. Sports Exerc.
18. Opplinger, R. A., D. H. Neilsen, J. E. Hoegh, and C. G. Vance. Bioelectrical impedance prediction of fat-free mass for high school wrestlers validated. Med. Sci. Sports Exerc.
(Abstract) 23:573, 1991.
19. Rising, R., B. Swinburn, K. Larson, and E. Ravussin. Body composition in Pima Indians: validation of bioelectrical impedance. Am. J. Clin. Nutr.
20. Ross, R., L. Leger, P. Martin, and R. Roy. Sensitivity of bioelectrical impedance to detect changes in human body composition. J. Appl. Physiol.
21. Scheltinga, M. R., D. O. Jacobs, T. D. Kimbrough, and D. W. Wilmore. Alterations in body fluid content can be detected by bioelectrical impedance analysis. J. Surg. Res.
22. Segal, K., Use of bioelectrical impedance analysis measurements as an evaluation for participating in sports. Am. J. Clin. Nutr.
66(Suppl. 3):469S-471S, 1996.
23. Segal, K. R., B. Gutin, E. Presta, J. Wang, and T. B. Van Itallie. Estimation of human body composition by electrical impedance methods: a comparative study. J. Appl. Physiol.
24. Segal, K. R., M. Van Loan, P. I. Fitzgerald, J. A. Hodgdon, and T. B. Van Itallie. Lean body mass estimation by bioelectrical impedance analysis: a four site cross-validation study. Am. J. Clin. Nutr.
25. Siri, W. E. Body composition from fluid spaces and density. In: Techniques for Measuring Body Composition.
J. Brozek and A. Henschel (Eds.) Washington, DC: National Academy of Sciences, 1961, pp. 223-244.
26. Vehrs, P. R., J. D. George, and C. Payne. Reliability of four methods of body composition assessment. Med. Exerc. Nutr. Health
27. Wilmore, J. H., A. Vodak, R. B. Parr, R. N. Girandola, and J. E. Billing. Further simplification of a method for determination of residual lung volume. Med. Sci. Sports Exerc.
28. Wilmore, K. M., P. J. McBride, and J. H. Wilmore. Comparison of bioelectrical impedance and near-infrared interactance for body composition assessment in a population of self-perceived overweight adults. Int. J. Obes.