点击显示 收起
1 From the Division of Nutrition, Department of Biomedicine and Surgery, University of Linkoping, Linkoping, Sweden
2 Supported by the Swedish Research Council (project no. 12172), the Swedish Nutrition Foundation, the County Council of Ostergotland, Knut and Alice Wallenberg's Foundation, Magnus Bergvall's Foundation, and Dr P Hakansson's Foundation. 3 Reprints not available. Address correspondence to E Forsum, Division of Nutrition, Department of Biomedicine and Surgery, University of Linkoping, SE-58185 Linkoping, Sweden. E-mail: elifo{at}ibk.liu.se.
ABSTRACT
Background: Knowledge of the biological variability of the hydration factor (HF), ie, the ratio between total body water and fat-free mass, is important when calculating total body fat by means of the commonly used two-component model, which is based on estimates of body weight and total body water. The effect of pregnancy on the biological variability of HF, and consequently on the precision of the two-component model, is unknown.
Objective: Our goal was to assess the effect of pregnancy on HF and its biological variability.
Design: HF was assessed in 33 women planning pregnancy and in 17 of these women during gestational weeks 14 and 32 and 2 wk postpartum. HF was calculated by using estimates of body weight, total body water obtained by means of deuterium dilution, and body volume measured by using underwater weighing.
Results: In the 17 women who became pregnant, HF was 0.718 ± 0.023, 0.723 ± 0.031, 0.747 ± 0.017, and 0.734 ± 0.020 before pregnancy, in gestational week 14, in gestational week 32, and 2 wk postpartum, respectively. The biological variability represented 2% of average HF in the nonpregnant state. The corresponding figure was >3% in gestational week 14 but 1.7% in gestational week 32.
Conclusion: The two-component model for assessing body fat is as appropriate during late gestation as it is in the nonpregnant state, although its precision may be impaired when applied during the first part of pregnancy.
Key Words: Biological variability • body composition • body density • hydration of fat-free mass • methodologic error • total body water • pregnancy
INTRODUCTION
The ratio of total body water (TBW) to fat-free mass (FFM) of the body, also called the hydration factor (HF), is surprisingly stable in mammals, an observation with important implications for body-composition studies, because the total body fat (TBF) of a living subject can be estimated from body weight and TBW if HF is known and constant (1). Such a two-component model is widely used in humans to assess TBF, a variable of fundamental importance when evaluating the nutritional status of groups and individuals. As indicated by Wang et al (1) and by Schoeller (2), earlier studies showed that the HF for human adults is 0.73. Additional studies reporting an average HF have been published for healthy adults (3-5), for infants and children (6), for old persons (4), and for pregnant women (7-9). However, as recently pointed out by Wells et al (10), there is a certain interindividual or biological variability in HF that affects the precision of the two-component model. Those authors therefore studied this variability in children aged 8–12 y and found it to represent only 2% of average HF, which they considered to be small enough for this model to be acceptable when assessing the TBF of children in this particular age group (10). The observations of Wells et al (10) warrant assessing the biological variability of HF in other groups of humans as well. We found only one study in which the biological variability in HF was studied in women of reproductive age (3). Furthermore, although it is known that HF increases during pregnancy (7-9), when women accumulate variable amounts of TBW, it is not known how pregnancy affects the biological variability of HF. van Raaij et al (9) and Kopp-Hoolihan et al (11) have expressed concern that the two-component model may not be appropriate when TBF is calculated during gestation.
The magnitude of HF can be assessed by using currently available body-composition methodology. However, all methods are associated with experimental errors, and when assessed in a group of subjects, the total variability in HF is therefore a function of these errors as well as of the biological variability in HF. These 2 components of the observed variability can be assessed by using propagation of error analysis. Application of such analysis to estimates of HF in pregnant women can give information regarding the usefulness of the two-component model during pregnancy. So far, no study has assessed the effect of gestation on HF through the use of a longitudinal design and by including measurements taken before, during, and after pregnancy. Additionally, the effect of pregnancy on the biological variability in HF has not been studied previously. The aims of the present study were therefore to assess HF and its biological variability in healthy women of reproductive age and to assess the effect of pregnancy on HF and its biological variability.
SUBJECTS AND METHODS
Subjects
Thirty-three healthy women planning pregnancy, living in the Linkoping area, and recruited by means of advertisements in the local press or through the health care system participated in the baseline measurements. Seventeen of these women became pregnant after 8–524 d (pregnancy group) and were studied again during gestational weeks 14 and 32 and 2 wk postpartum. Gestational age was estimated on the basis of an ultrasound scan taken during gestational week 12–14. Each of the women in this group remained healthy during pregnancy (no proteinuria, generalized edema, preeclampsia, or eclampsia) and delivered one healthy, full-term baby. The women's body weights before childbirth were recorded in the delivery room. The remaining 16 women (nonpregnancy group) did not become pregnant in time to be studied during pregnancy. The study was approved by the ethics committee of the University of Linkoping.
Protocol
The following procedure was performed when pregnancy was planned, during gestational weeks 14 and 32, and 2 wk postpartum. After an overnight fast, the subject arrived at the hospital and body weight, body volume (BV), and body density were measured and an oral dose of deuterium was administered. During the following 14 d, the subject collected urine samples for the estimation of TBW.
Deuterium dilution
Each subject was given an accurately weighed dose of 2H2O (0.05 g/kg body wt) after the collection of 2 or 3 background urine samples during a period of 2–7 d before dosing. Another 5 urine samples were collected 1, 4, 8, 11, and 15 d after dosing. The date and time of day when a sample was collected were always noted. Urine samples were stored in glass vials with internal aluminum-lined screw caps at 4 °C until sample collection was completed, after which they were stored at –20 °C until analyzed. Deuterium enrichments of the dose and urine samples were analyzed by using an isotopic ratio mass spectrometer fitted with a CO2/H2/H2O equilibrium device (Deltaplus XL; Thermoquest, Bremen, Germany). The procedure described by Thielecke and Noack (12) was followed, except that the equilibration time was 180 min. Deuterium dilution space (ND) was calculated from zero-time enrichment obtained from the exponential isotope disappearance curve. The mass spectrometric response was standardized by using Vienna standard mean ocean water. Dose and urine samples from each subject and measurement occasion were always analyzed simultaneously within the same equilibrium device, when a linear mass spectrometric response was also confirmed. Analytic precision in the measurement range for results expressed as a mole fraction was 0.44 ppm. TBW was calculated as ND/1.04.
Underwater weighing
Each subject was weighed in air without clothes (KCC 150; Mettler-Toledo, Albstadt, Germany) after she had emptied her bladder. The precision of this estimate was 0.01 kg. The subject then entered a water tank that was equipped with an appropriate balance (SB16001; Mettler-Toledo), and 2 or 3 weight belts were attached to her body. Thereafter, she was connected to a spirometer (Volugraph 2000, Siemens-Elma, Stockholm) based on the closed-circuit helium dilution principle (13) that displayed her breathing movements. When she felt comfortable, the subject submerged herself slowly under water to a lying position. Weight under water was recorded 7 times. The subject was holding her breath and remained motionless when her underwater weight was recorded. Each time when this weight was recorded, lung volume was measured, exactly at the time of weighing. The water temperature was measured immediately before the procedure, which lasted for 5 min. Each subject practiced the procedure (without using helium) once before the measurements were taken. BV was calculated as follows (14):
RESULTS
Subjects
Nonpregnant state
The characteristics of the subjects at the time when pregnancy was planned are shown in Table 1. Although the women varied considerably with respect to body weight, BMI, TBW, BV, TBF, and FFM, there were no significant differences between the pregnancy and nonpregnancy groups. Sixteen of the 33 women in the study were in the preovulatory phase and 17 were in the postovaluatory phase when they were measured before pregnancy. Corresponding figures for the women in the pregnancy group were 11 and 6.
View this table:
TABLE 1. Characteristics of the subjects at the time pregnancy was planned1
Effect of pregnancy
The women in the pregnancy group (n = 17) gained 16.7 ± 6.4 kg (range: 8.1–25.3 kg) of body weight during the entire gestation. Their body weight, TBW, BV, TBF, and FFM throughout the reproductive cycle are shown in Table 2. During gestational week 14, average body weight, TBW, BV, TBF, and FFM were 1.8 kg, 1.0 kg, 1.8 L, 0.8 kg, and 1.0 kg higher, respectively, than before pregnancy, but these increases were not significant. However, in gestational week 32, the women had gained significant amounts of body weight (10.7 kg), TBW (6.6 kg), TBF (3.6 kg), and FFM (7.1 kg) when compared with the corresponding values before conception. Their BV had also increased significantly by 10.9 L in gestational week 32 compared with the prepregnant value. Furthermore, at the measurement 2 wk postpartum, body weight, TBW, BV, FFM, and TBF, in kg as well as in %, were all significantly higher than before pregnancy.
View this table:
TABLE 2. Body weight, total body water (TBW), body volume (BV), fat-free mass (FFM), and total body fat (TBF) of women in the pregnancy group at different stages of reproduction1
Hydration factor
Nonpregnant state
As shown in Table 3, average HF at the time when pregnancy was planned was 0.720 for all women in the study, 0.718 for the women in the pregnancy group, and 0.722 for the women in the nonpregnancy group. HF was not correlated with day of menstrual cycle (n = 33).
View this table:
TABLE 3. Average hydration factor (HF) and total variability in HF in women planning pregnancy, the effect of pregnancy on the mean and total variability in HF, and the contributions to total variability of methodologic error and biological variability with the use of 2 sets of assumptions
Effect of pregnancy
HF at different stages of reproduction is also shown in Table 3. For women in the pregnancy group, HF during gestational week 14 was 0.723 and was not significantly different from the value obtained for these women before conception. During gestational week 32, HF was 0.747, which was significantly higher than that during gestational week 14 or before pregnancy. At the measurement 2 wk postpartum, HF was 0.734, which was significantly higher than that before conception.
Variability in the hydration factor
Nonpregnant state
The results obtained for all women in the study at the time when pregnancy was planned are shown in Table 3. The SD of the women's average HF was 0.021. Propagation of error analysis with the use of 2 sets of assumptions indicated that the fraction of the total variability that could be attributed to biological variability was 41–59%. The implication is that biological variability (0.013–0.016) represented 1.8–2.2% of average HF. The corresponding results obtained for the nonpregnancy and pregnancy groups were 1.2–1.7% and 2.3–2.7%, respectively.
Effect of pregnancy
Also shown in Table 3 are the results obtained in the pregnancy group during and after pregnancy. During gestational week 14, total variability in HF was as high as 0.031. Propagation of error analysis showed that biological variability then represented 73% and 82% of total variability with use of the first and second sets of assumptions, respectively. Thus, biological variability represented 3.6–3.9% of average HF. During gestational week 32, biological variability represented 5% and 53% of total variability, respectively, when the 2 sets of assumptions were used. The total variability was as low as 0.017 in gestational week 32, and thus biological variability then represented 0.5–1.7% of average HF. The total variability 2 wk postpartum was 0.020, and with use of the 2 sets of assumptions, biological variability was then found to represent 29% and 59%, respectively, of total variability. Consequently, biological variability represented 1.4–2.1% of average HF 2 wk postpartum. Thus, using the first as well as the second set of assumptions, the propagation of error analysis showed that the effect of pregnancy on the biological variability in HF was largest in week 14 and smallest in week 32.
DISCUSSION
On average, the 17 pregnant women we studied gained 10.7 kg of body weight by gestational week 32 and as much as 16.7 kg during the complete gestation. For 10 of these women, weight gain was above that recommended by the Institute of Medicine (19). Furthermore, these women gained almost 7 kg of TBW during the first 32 wk of pregnancy, which is somewhat high when compared with data published by Hytten (20). The results of earlier studies suggest that, when compared with other Western women, Swedish women tend to have high weight gains during pregnancy (21, 22), but no recent comprehensive data on this topic are available. Therefore, we cannot definitely conclude that the women we studied are typical of Swedish women in general.
The present results with respect to average HF for healthy women of reproductive age can be compared with those published by Fuller et al (3) and Visser et al (5), who found HF to be 0.738. Hewitt et al (4) reported slightly lower values ranging from 0.707 to 0.722 depending on the method used. Using data provided by Siri (15), Fuller et al (3) calculated HF to be 0.719, whereas 0.73 represents an average estimate of HF in adults (1, 2). Our figure, 0.72, is thus in good agreement with previously published results.
Our estimate of HF during gestational week 14 did not differ significantly from the prepregnant value, whereas a significant increase to 0.747 was found during gestational week 32. The value at week 32, however, is lower than several previous estimates of HF at this stage of gestation (7, 23, 24) and is 0.015 units lower than the figure reported by Catalano et al (7) for women in gestational week 30 (0.762). The reason for this discrepancy between the 2 studies is not obvious, but certain methodologic differences may be relevant. In our study, TBW was calculated by means of the back-extrapolation method, whereas Catalano et al (7) used the plateau method. This difference in methods may explain about one-third of the difference in average HF between the 2 studies. Furthermore, when oxygen-18 is used to assess TBW, as in the study by Catalano et al (7), the fact that oxygen-18 overestimates TBW by 1% is usually taken into account (2). However, it is unclear whether Catalano et al (7) made this correction, thereby making it difficult to explain the discrepancy between the 2 studies. This comparison illustrates the importance of considering the procedures used to estimate TBW when assessing the magnitude of HF. It is also of interest to compare our results with those published by van Raij et al (9), who estimated that women increase their average HF from 0.724 before pregnancy to 0.742 during gestational week 32, ie, an increase of 0.018 units. The corresponding increase in our study was 0.029 units.
Expressing the estimated biological variability as a percentage of average HF gives a figure of 2% for our nonpregnant subjects, which is similar to the 2.1% reported by Wells et al (10) for children 8–12 y old, which indicates that the two-component model is also appropriate for healthy adult women. With respect to pregnancy, the largest biological variability was found during gestational week 14, which was probably an effect of the adaptive changes in circulation and body composition typical of the first part of pregnancy. On the other hand, biological variability was comparatively small during gestational week 32, when it was only 1.7% of average HF. Thus, biological variability appears to be lower during gestational week 32 than during the nonpregnant state. We want to point out, however, that such a decrease cannot really be statistically confirmed. Although it is possible to test whether the total variability during gestational week 32 is significantly lower than the corresponding value before pregnancy, the results of such a test are difficult to interpret because the relative contributions of methodologic error and biological variability may differ for the 2 measurements. Therefore, a statistically significant difference in total variability cannot be interpreted in terms of a statistically different biological variability. However, our finding of comparatively low biological variability during gestational week 32 is of interest in relation to the finding by Catalano et al (7) that the total variability during gestational week 30 represented only 1.6% of average HF. Considering the methods used by these authors, such a low figure for total variability can only be obtained if the biological variability of HF is quite low. Thus, it seems reasonable to conclude that a two-component model based on assessments of TBW and body weight is a satisfactory means of assessing TBF in women at this stage of pregnancy. It should be noted, however, that this conclusion may not be valid for pregnant women with generalized edema.
Our findings warrant some comments regarding the magnitude of the methodologic error when estimating body weight, TBW, and BV. As pointed out by Fuller et al (3), the contribution of the error in the assessment of body weight to the methodologic error in HF estimated with the use of a three-component model is negligible. Those authors also noted that the error in estimates of TBW is unimportant for the methodologic error of HF, because the errors from this source largely cancel out during propagation of error analyses. Thus, our results would have remained nearly the same had we, as proposed by Murgatroyd and Coward (17), selected 2% instead of 1.05% as the error associated with measurements of TBW. Instead, the most important factor contributing to the methodologic error in HF is the error when estimating BV. Several authors have assessed the error associated with the assessment of body density, and the reported values range from 0.0011 to 0.0043 g/cm3 (25). It may be argued that our error, 0.0016 g/cm3, was not assessed in a satisfactory way, because it was based on only 4 observations. However, an underestimation of this error would have resulted in an even smaller biological variability in HF. If, on the other hand, we overestimated the error of the underwater weighing measurement, ie, if the true value for this particular error is on the order of magnitude of only 0.0011 g/cm3 (25, 26), our results would have remained very similar. Thus, using 0.0011 g/cm3 instead of 0.0016 g/cm3 in a propagation of error analysis with all 33 women planning pregnancy and the second set of assumptions in Table 3 would change the biological variability only from 0.016 to 0.018 and increase the biological variability from only 2.2% to 2.6% of average HF.
Another important consideration is related to possible alterations in methodologic error, especially in the underwater weighing technique, throughout our study. This error could have decreased as the participants became more proficient with the technique. This error could also have increased if the pregnant state made the measurements more difficult for the subjects. To obtain the best possible assessments of BV, the women practiced the underwater weighing procedure before the real measurements were carried out. Therefore, they were comfortable with the procedure already when measured before pregnancy. We have no indications whatsoever that pregnancy made it more difficult for the women to participate in the underwater weighing procedure. Thus, we do not believe that the longitudinal design of our study caused any important changes in the magnitude of the methodologic error associated with this technique.
An additional factor of interest when calculating the biological variability in HF during pregnancy is whether the errors in the different methods are expressed in % or in kg and L. Our results in Table 3 show that, irrespective of how these errors are expressed, estimates of biological variability correspond to an SD of 0.02 before, during, and after pregnancy, with the exception of gestational week 14. At this stage of pregnancy, biological variability corresponded to a higher SD (>0.026) irrespective of how the error was expressed.
The present article reports that the biological variability in HF is 2% of its average value in nonpregnant women and that this figure increases during early pregnancy but decreases during late gestation. These findings support the conclusions that the two-component model is as appropriate during late gestation as it is in the nonpregnant state, whereas it may be associated with a lower precision during the first part of pregnancy.
ACKNOWLEDGMENTS
We thank the all the women who participated in this study. We also thank Hanna Olausson for valuable help during data collection and Karin Boström for help in recruiting women planning pregnancy. Special thanks are due to Olle Eriksson for providing the method of calculation needed to perform the propagation of error analysis.
EF designed this study and ML had the primary responsibility for recruiting and investigating the subjects and for laboratory and data analysis. Both authors contributed to the preparation of the manuscript and neither had any conflict of interest.
REFERENCES