- Email: [email protected]
Longitudinal changes in the relationship between body mass index and percent body fat in pregnancy
Longitudinal Changes in the Relationship Between Body Mass Index and Percent Body Fat in Pregnancy CAROL A. LINDSAY, MD, LARRAINE HUSTON, MS, SAEID B. AMINI, PhD, MBA, AND PATRICK M. CATALANO, MD Objective: To determine the longitudinal relationship between body mass index (BMI) and percent body fat in w o m e n before and during pregnancy. Methods: Twenty-seven healthy, nonobese w o m e n were evaluated before conception, in early gestation (12-17 weeks), and in late gestation (33-36 weeks). Height and weight were measured and BMI was calculated. Percent body fat was estimated using hydrodensitometry with correction for residual lung volume. Results: The correlation between BMI and percent body fat before conception was r = 0.693 (P < .005); in early gestation it was r = 0.723 (P < .005) and in late gestation r = 0.633 (P < .005). The mean pregravid BMI was 21.54 and the 95% predictive confidence interval (CI) for percent body fat was 18.2, 26.5%. For the mean BMI of 22.26 in early gestation, the predictive 95% CI for percent body fat was 20.0, 29.0%. In late gestation, the mean BMI was 26.04 with a predictive 95% CI for percent body fat 22.5, 30.8%. Conclusion: In nonobese w o m e n the correlation between BMI and percent body fat remains significant during pregnancy, although the 95% CI for predicting percent body fat from the mean BMI ranges widely. (Obstet Gynecol 1997;89: 377-82. Copyright © 1997 by The American College of Obstetricians and Gynecologists.)
Body mass index (BMI), defined as weight (kg)/height (m) 2, has been used as an estimate of obesity in both pregnant and nonpregnant individuals. Additionally, BMI has been used to predict various pregnancy outcomes and served as the basis for recommendations regarding appropriate weight gain and caloric intake during pregnancy. ~ Because BMI has been applied in both research and clinical settings as an indicator of obesity in pregnancy, 2'3 it is important to know how From the Department of Reproductive Biology, Case Western Reserve University at MetroHealth Medical Center, Cleveland, Ohio. Supported by NIH HD22965(PMC) and the General Clinical Research Center RR 00080.
VOL. 89, NO. 3, MARCH 1997
well this measure estimates percent body fat during pregnancy. The range for the correlation coefficient of BMI with percent body fat in nonpregnant w o m e n is between .60 and .82. 4 Based on our review of the literature using a MEDLINE search (1969-1996), the relationship between BMI and percent body fat has not been evaluated during pregnancy. Hence, this study was done to evaluate prospectively the longitudinal relationship between BMI and percent b o d y fat in nonpregnant w o m e n who subsequently conceived and to determine whether the relationship changed significantly over time. We hypothesized that BMI would be a poorer indicator of percent body fat in pregnancy than in the nonpregnant individual due to the relative increases in fat-free mass (such as plasma volume, placental mass, amniotic fluid (AF), and uterine mass) and that the strength of the correlation between percent body fat and BMI would decrease with advancing gestation.
Materials and Methods Our sample consisted of 27 healthy, multiparous, white, nonsmoking, nonobese w o m e n selected from a population of nonpregnant w o m e n with normal glucose tolerance who were contemplating pregnancy. Sixteen subjects participated in longitudinal studies 5 of glucose metabolism published previously. The enrollment period for this study was approximately 1 year. Pregravid studies were repeated every 6 months until the w o m e n conceived, with the measurements immediately preceding the pregnancies used as the baseline pregravid measurements. We recruited approximately 10-20% of subjects invited to participate in this study. Each subject had a pregravid BMI below 29.0. Mean maternal age at delivery was 32.3 years (range 23-40). Mean parity was 2.1 (range 1-3), and 89% of the subjects had at least a
0029-7844/97/$17.00
PI1 S0029-7844(96)00517-0
377
high school education. None of the women had medical problems, prior bowel surgery, or large variations in pre-pregnancy weight. Each subject had three estimates of body fat: pregravid, early gestation (defined as 12-17 weeks), and late gestation (defined as 33-36 weeks). Gestational age was determined by last menstrual period and confirmed in all subjects by first-trimester ultrasound. At each study period subjects were weighed on a calibrated scale and their heights were measured while the subjects stood with their backs against a wall. Body composition was estimated using underwater weighing or hydrodensitometry in the general clinical research center. Each subject was weighed underwater and the bathing suit weight was corrected/~ Residual lung volume was determined simultaneously by the nitrogen or helium washout technique. The subject breathed room air during the underwater weighing, followed by administration of 100% oxygen and measurement of the change in nitrogen or helium concentration of the expired air over 5-7 minutes. We used this value to calculate residual lung volume." Total body water increases by 0.7 L in early gestation, as d e m o n s t r a t e d by Lukaski et al. s The multicompartment model of body composition requires direct measurements of the various components of fatfree mass (bone mineral, non-bone mineral, water, and protein). 9 Theoretical equations derived to calculate body fat in pregnancy account for the increase in total body water, ~° but none have been tested in a prospective manner in early gestation. Therefore, in the current study the two-compartment model was used with equations derived in nonpregnant women before conception and in early gestation. Body fat was determined with the Keys and Brozek 11 equation (body fat = 420.1 divided by body density minus 381.3) for the pregravid and early gestation subjects. In late gestation there is an increase in total body water as measured with 2H21so.~2 During late gestation, body fat was estimated based on an equation derived at our institution by Catalano et al ~2(body fat = 518.57 divided by body density minus 476.30). This equation adjusts for increase in total body water by using a hydration constant of 0.76. j2 Percent body fat was calculated as body fat divided by total body weight. Once these data were collected, regression equations were created with BMI as the independent variable and percent body fat the dependent variable. Higher-order polynomial regression did not yield an improvement over simple linear regression; therefore, linear regression was used. The correlation between the two continuous variables was measured by Pearson's correlation
378 Lindsay et al
BM[ rout Percent Body Fat
Table 1. Body Composition at Each Period
Weight (kg) BMI (kg/m2) Body fat (%) Fat mass (kg) Fat-free mass (kg)
Pregravid
Early gestation
59.7 = 6.4 21.5 z 2.4 22.3 + 5.1 13.5 + 4.0 46.2 -+4.3
61.7 -+ 7.2 22.5 -+ 2.8 24.4 -+ 5.0 15.2 -+4.4 46.4 + 4.5
Late gestation 73.3 + 8.3* 26.0 ÷ 2.6* 26.6 -+ 4.8* 19.4 -+4.8 52.9 _+5.7
BMI = body mass index. Data are expressed as mean _+one standard deviation. *Changes over time significant by analysis of variance with repeated measures (P < .001). Changespregravid versus early gestation, pregravid versus late gestation, and early gestation versus late gestation also significant by Scheffetest, with an overall P < .05.
coefficient. In early gestation, the correlation between BMI and percent body fat, as determined by the Keys and Brozek ~1 equation, did not differ significantly from that determined by the equation derived by Catalano et al, ~2 the Siri 13 equation, or the equations by van Raaij et al ~° for 10 and 20 weeks' gestation; therefore, the Keys and Brozek 11 equation was used in the analysis of early gestation. The 95% confidence interval (CI) for predicting percent body fat was determined for the mean, upper, and lower extremes of BMI in each study period. Finally, an analysis of variance with repeated measures was done for each subject's change in weight, BMI, and percent body fat during pregnancy; and a secondary analysis using the Scheffe method of contrasts determined the difference in outcome measures at specific points. The value considered significant for this study was adjusted to P < .005 (Bonferroni's correction) due to the multiple comparisons made. The statistical software Statview 4.02 (Abacus Concepts, Berkeley, CA) was used for some statistical analyses.
Results Pregravid, early gestation, and late gestation weight; BMI; and percent body fat are listed in Table 1. Mean weight gain during early gestation was 1.96 kg, 87% of the increase in weight being fat. Late in gestation the percentage of weight gain due to the accumulation of fat decreased. Mean weight gain from pre-pregnancy until late gestation was 12.61 kg, and the percentage of the total weight gain in late gestation due to fat was only 47%. The correlation coefficient between pregravid BMI and percent body fat was r = 0.69 (Y = -9.491 + 1.478X, 95% CI 0.19, 0.76, P < .005). Given a mean pregravid BMI of 21.5, the predictive 95% CI for percent body fat is between 18.2 and 26.5%. The predictive 95% CI was also calculated at the extremes of BMI. For a BMI of
Obstetrics & Gynecology
18.1, the 95% predictive CI for percent body fat was between 12.8 and 21.7%. At the upper extreme, a BMI of 27.5 pregravid gave a predictive 95% CI between 26.0 and 36.2% body fat. The correlation between BMI and percent body fat in early pregnancy was r = 0.72 (Y = -11.762 + 1.63X, 95% CI 0.47, 0.86, P < .005). The difference between the correlation coefficient of BMI and percent body fat pregravid and in early gestation was not statistically significant by analysis of variance of repeated measures (P = .83). For a mean BMI of 22.3 in early gestation, the predictive 95% CI for percent body fat was between 20.0 and 29.0%. When measured in early gestation, a BM1 of 18.4 gave a 95% predictive CI for percent body fat between 13.5 and 23.1%. A BMI of 29.4 gave a 95% predictive CI for percent body fat between 30.5 and 41.7%. The correlation between BMI and percent body fat decreased in late pregnancy to r = 0.63 (Y = -3.735 + 1.166X, 95% CI 0.333, 0.817, P < .005). Neither the difference between the correlation coefficients of BMI and percent body fat pregravid and in late gestation nor the difference between the correlation coefficients in early and late gestation when compared by analysis of variance of repeated measures was statistically significant (P = .71 and .56, respectively). In late gestation, the mean BMI was 26.0 with a predictive 95% CI for percent body fat between 22.5 and 30.8°/,,. At the extremes, a BMI of 21.9 gave a 95°/,, predictive CI for percent body fat between 17.3 and 26.4%. At the u p p e r extreme, a BMI of 31.7 gave a 95% predictive CI for percent body fat between 28.3 and 38.1%. In all periods studied, the 95% CI for predicting percent body fat from BMI increased at the u p p e r extreme. The longitudinal changes in weight, BMI, and percent body fat with advancing gestation were all statistically significant (P < .001). All pair-wise comparisons of changes in weight, BMI, and percent body fat between pregravid and early and late gestation were statistically significant (P < .05). Finally, the ability to estimate percent body fat using weight alone was relatively poor compared with BMI, weight and percent body fat correlation coefficient r = 0.51 (P - .006), r = 0.54 (P .003), and r = 0.41 (P = .034) (pregravid, early gestation, and late gestation, respectively).
Discussion Our results indicate that the correlation between BMI and percent body fat in early pregnancy is consistent with that reported in the literature for nonpregnant subjects. However, although statistically significant, in early gestation BMI explains 52% of the variance in
VOL. 89, NO. 3, MARCH 1997
percent body fat and in late gestation it explains only 40% of the variance in percent body fat. Body composition changes during pregnancy. Of the 12.5-kg weight gain, approximately 3.5 kg is due to the accumulation of fat, the accretion of fat occurring predominately in the first half of pregnancy. In our sample, 29°/`, of the fat accumulation occurred during early gestation, whereas only 16% of the total weight gain occurred during this period. Products of conception represent 10, 20, 30, and 40% of the total weight gain at 10, 20, 30, and 40 weeks' gestation, respectively. ~4 This is consistent with our findings of decreasing amounts of total weight gain due to accretion of fat in late gestation. Percent body fat varies widely with the method used for its determination. Lederman et a115 found percent body fat to range from 21.5 as determined by total body water with dilution with D20, 22.3 with hydrodensitometry, and 29.1 by total body potassium in the same subject at 14 weeks. This is consistent with our findings of 24.4 percent body fat in early gestation as determined by hydrodensitometry. The same group also found a wide range in late gestation (defined as 37 weeks): 21.8 by total body water, 28.0 by hydrodensitometry, and 37.2 by total body potassium. Is Catalano et a112 also found a wide range in the third trimester: 34.4 by hydrodensitometry and 26.5 by total body water. Forsum et al 1~ used skinfolds to estimate percent body fat in two different groups of patients. In early gestation, percent body fat was 32.0 and 29.8, at 30 weeks it was 31.7 and 29.6, and at 36 weeks it was 30.2 and 28.3%. ~ In our study, percent body fat in late gestation as determined by hydrodensitometry was 26.6. The wide ranges of percent body fat are a reflection of violations of the many assumptions that must be made to calculate percent body fat with each different method in pregnant w o m e n as well as individual variations in the patients. Hydrodensitometry is one of the conllTlon research methods used in estimating body composition in humans and frequently is used as the standard by which other methods are evaluated. Hydrodensitometry, however, is an indirect measure of body composition. The main determinants of human body density are fat and bone. The density of bone is approximately 3.0, adipose tissue (which contains 83% fat, 15% water, and 2% protein) r'~s is 0.94, water is 0.996, and all other tissues are approximately 1.060. Fat appears to be the main factor affecting the specific density of an individual. The density of body fat is 0.900 g / m L , and all other body tissues, ie, fat-free mass, have a mean density of 1.100 g / m L ; therefore, low values of density indicate increasing obesity and greater values denote leanness. For the purposes of estimating body composition, the h u m a n body can be divided into components. Hydro-
Lindsay et at
BMI nnd Percent Body Fat
379
densitometry relies on a two-compartment model, which divides the body into fat and fat-free mass. Throughout gestation the amount of body water increases. In a longitudinal study of 15 women, Lukaski et al s found total body water to increase from a prepregnancy mean of 31.7 L to 32.2 L at 14-16 weeks and 38.8 L at 36-38 weeks. In a two-compartment model, an increase in total body water leads to a decrease in total body density. The density of water is approximately 0.996 and the density of fat-free mass is approximately 1.100; therefore, an increase in total body water leads to a decrease in density and an overestimation of body fat 12 when density equations are used for nonpregnant w o m e n during gestation. Keys and Brozek ~1 derived an equation for estimating body fat (420.1 divided by body density minus 381.3). Non-gravid b o d y density calculations assume a hydration constant between 0.72 and 0.73 of the free-fat mass. Although there is an increase in total body water in early gestation, no equations to estimate percent body fat from density have been tested in a prospective manner in early gestation. Because there is an increase in maternal total body water in late gestation, the use of equations with a hydration constant of 0.72-0.73 m a y result in overestimation of fat mass. Recently, an equation was derived for late gestation by Catalano and colleagues 12 (body fat = 518.57 divided by body density minus 476.30) that derived a hydration constant of 0.76 based on measurement of both density and total body water using 2H2180. Several assumptions apply with hydrodensitometry in estimating body fat. Hydrodensitometry as a measure of percent body fat assumes a constant hydration and a constant proportion of bone mineral to muscle in the fat-free body, 19 not varying by more than 0.004. 20 It also assumes that there is minimal air in the gut. It is not possible to measure air in the gut, and this causes an error of approximately 1.5% in the estimation of body fatJ 9 Finally, error also can be introduced from the measurement of residual lung volume, causing an error of _+0.003, which can be minimized with repeated determinations. These assumptions are used in nonpregnant w o m e n and we assume they apply to pregnant women as well, with the exception of the increase in total body water as noted previously. There is also a decrease in residual volume of approximately 200 mL in pregnancy, but the residual lung volume is measured at each density determination; therefore, it is assumed that this difference is accounted for in the density measurement. Although hydrodensitometry is probably the best method for determining the contribution of fat mass to weight in pregnancy, there are also potential technical
380 Lindsay et al
BMI and Percent Body Fat
difficulties in performing hydrodensitometry during pregnancy. Subjects must be highly motivated, cooperative, and able to remain submerged and motionless for the amount of time required to obtain the measurement. 17 The technique requires special equipment and about 1 hour to obtain the subject's density with residual lung volume measurement. In addition, the estimation of body composition by hydrodensitometry in pregnancy includes fetal mass, placental mass, and AF. The major methodologic disadvantage to the use of hydrodensitometry in pregnancy is that the estimation of percent body fat is dependent on the density of fat-free mass. Density of fat-free mass decreases in pregnancy due to an increase in total body water; therefore, fat mass will be overestimated using standard, nongravid equations. Hydrodensitometry is also expensive and time consuming and requires personnel skilled in the measurements and a cooperative patient. It is not practical for widespread application. Body mass index is used widely as an index of obesity. Obesity is defined as an increase in the percent of body fat I that has been estimated based on BMI by various organizations and investigators. A good measure of obesity should correlate highly with relative adiposity and be independent of height. 21 A number of body mass indices have been derived, including the weight-to-height ratio (W / H), Quetelet index (W / H 2), ponderal index (W1/3/H), Sheldon index (H/W1/3), Abde-Malek index (cWI2/H 3"3, where c = 4 × 106 for w o m e n and height is in centimeters), and percent desirable body weight expressed as percentage of the mean weight for a given height and sex. There was a strong intercorrelation between the various BMIs (ranging from 0.82 to 1.0). These indices are correlated strongly with weight and weakly with height. All BMIs correlated strongly with percent body fat as determined by densitometer (r = 0.87-0.92) in a selected nonpregnant female population. The correlation reported in the general population was not as strong (0.60-0.82). Quetelet index was found to be the most stable in young adults 22 and was used in the present study. The 95% C1 for the percent body fat as estimated from BMI was found to be wide, however, ranging from 10-12%. 4 This range was consistent with our findings in the current study. The 95% CI for predicting percent body fat from BMI was widest at the upper extreme of BMI both before gestation and during pregnancy. This would imply that BMI is less precise at predicting percent body fat at the u p p e r extremes of BMI. There are several pitfalls in the use of BMI as a measure of obesity. First, BMI is not completely independent of stature; the correlation coefficient between BMI and stature is approximately 0.12. Second, BMI is
Obstetrics & Gynecology
influenced nearly equally by the lean a n d fat compon e n t s of the body. 23 Two i n d i v i d u a l s can have the same BMI a n d still vary w i d e l y in the percent b o d y fat because muscle a n d fat contents are n o t necessarily correlated. 24 It has b e e n proposed that, on an i n d i v i d u a l basis, BMI not be used as a m e a s u r e of obesity, a l t h o u g h it m a y be useful in p o p u l a t i o n studies. 4 A simple m e a s u r e of obesity in p r e g n a n c y w o u l d be helpful because certain complications of p r e g n a n c y are associated w i t h i n c r e a s i n g obesity, i n c l u d i n g preeclampsia, gestational diabetes mellitus, h y p e r t e n s i o n a n d cardiovascular disease, s h o u l d e r dystocia, rate of cesarean delivery, w o u n d infections a n d dehiscence, anesthesia risks, increased blood loss, a n d other postp a r t u m complications. There is also controversy regarding the weight gain in p r e g n a n c y that results in the best outcome. A n objective m e a s u r e of percent b o d y fat m a y help further clarify i n d i v i d u a l caloric requirements. Excessive weight gain in p r e g n a n c y even in the absence of obesity is associated with both delivery of a large for gestational age infant a n d increased rate of cesarean delivery even after controlling for birth weight. Also, excessive weight gain is more likely to be retained after the p r e g n a n c y . ~ At least in n o n o b e s e w o m e n , one can use BMI to estimate percent b o d y fat, which m a y be a better predictor of adverse outcome than weight alone; however, further s t u d y in this area is necessary. The strength of this s t u d y is that it was conducted l o n g i t u d i n a l l y on the same subjects, decreasing the variability d u e to i n d i v i d u a l differences. Application of its results are limited, however, because it was cond u c t e d in healthy, m u l t i p a r o u s , n o n o b e s e w o m e n . Therefore, the results m a y not apply to a dissimilar group. Correlations m a y differ in obese patients or in patients with disease processes that m a y alter b o d y composition, such as excess e d e m a formation. Also, b o d y composition changes t h r o u g h o u t gestation; therefore, equations m u s t be developed specific for each period of gestation to d e t e r m i n e percent b o d y fat. O u r results indicate that BMI can be used as a m e a s u r e of percent b o d y fat in n o n o b e s e p r e g n a n t w o m e n , d e m o n s t r a t e d by a significant correlation as estimated by h y d r o d e n s i t o m e t r y that was consistent with that reported for the n o n p r e g n a n t subject. This correlation is superior to the correlation of weight alone with percent b o d y fat. Moreover, this correlation of BMI a n d percent b o d y fat is better in the pregravid a n d early gestation patient, w h e n m u c h of the a c c u m u l a t i o n of fat is occurring. However, in late gestation w h e n the maternal weight gain is affected more by growth of fetal tissues, maternal water, p l a s m a v o l u m e , a n d AF volume, this correlation decreases and should be used with greater caution.
VOL. 89, NO. 3, MARCH 1997
Body mass index is easy to calculate, can be determ i n e d cheaply, a n d relies on i n f o r m a t i o n generally obtained at a n y initial health assessment. For this reason, it w o u l d be r e c o m m e n d e d for general application to d e t e r m i n e percent b o d y fat in a p o p u l a t i o n ; however, d u e to the wide CI for predicting percent b o d y fat, BMI m a y not be precise e n o u g h for use in the i n d i v i d u a l patient. It is i m p o r t a n t to note that this s t u d y was c o n d u c t e d in a n o n o b e s e p o p u l a t i o n a n d the results m a y not be applicable to obese w o m e n .
Refel'enc¢s 1. Ramin SM, Cunningham FG. Obesity in pregnancy. In: Cunningham FG, MacDonald PC, Gant NF, LevenoKJ, Gillstrap LC 1II,eds. Williams obstetrics. 19th ed. Norwalk, Connecticut: Appleton & Lange, 1995:l 14. 2. Wolfe HM, Sokol RJ, Martier SM, Zador IE. Maternal obesity: A potential source of error in sonographic prenatal diagnosis. Obstet Gynecol 1990;76:339-42. 3. Wolfe HM, Zador IE, Gross TL, Martier SS, Sokol RJ. The clinical utility of maternal body mass index in pregnancy. Am J Obstet Gynecol 1991;164:1306-10. 4. Smalley KJ, Knerr AN, Kendrick ZV, Colliver JA, Owen OE. Reassessment of body mass indices. Am J Clin Nutr 1990;52:405-8. 5. Catalano PM, Tyzbir ED, Wolfe RR, Calles J, Roman NM, Amini SB, et al. Carbohydrate metabolism during pregnancy in control subjects and women with gestationaI diabetes. Am J Physiol 1993;364:E60-7. 6. Akers R, Buskirk ER. An underwater system utilizing "force cube" transducers. J AppI Physiol 1969;26:64932. 7. Darling C, Cournand A, Richards DW. Studies on the intrapuhnonary mixture of gases. An open circuit method for measuring residual air. J Clin Invest 1940;19:609-18. 8. Lukaski HC, Siders WA, Nielsen EJ, Hall CB. Total body water in pregnancy: Assessment by using bioelectrical impedance. Am J Clin Nutr 1994;59:578-85. 9. Lederman SA. Recent issues related to nutrition during pregnancy. J Arn Coil Nutr 1993;12:91-100. 10. van Raaij JM, Peek ME, Vermaat-Miedema SH, Schonk CM, Hautvast JG. New equations for estimating body fat mass in pregnancy from body density or total body water. Am J Clin Nutr 1988;48:24-9. 11. Keys A, Brozek J. Body fat in adult man. Physiol Rev 1953;33:245 325. 12. Catalano PM, Wong WW, Drago NM, Amini SB. Estimating body composition in late gestation: A new hydration constant for body density and total body water. Am J Physiol I995;268:E153-8. 13. Siri WE. Body compositionfrom fluid spaces and density: Analysis of methods. In: BrozekJ, Henschel A, eds. Techniques for measuring body composition. Washington, DC: National Academy of Science-NationalResearch Council, 1961:223-44. 14. Hytten F, Chamberlain G, Weight gain in pregnancy. In: Hytten F, Chamberlain G, eds. Clinical physiology in obstetrics. 2nd ed. Oxford: BlackwellScientific Publishers, 1991:173--203. 15. Lederman SA, Pierson RN, Wang J, Paxton A, Thornton J, Wendel J, et al. Body composition measurements during pregnancy. Bas Life Sci 1993;60:193-5. 16. Forsm'nE, Sadurskis A, Wager J. Estirnationof body fat in healthy Swedish women during pregnancy and lactation. Am J Clin Nutr 1989;50:465 73.
Lindsay et al
BMI at~d Percent Body Fat
381
17. Garrow JS. New approaches to body composition. Am J Clin Nutr 1982;35:1152-8. 18. Jensen MD. Research techniques for body composition assessment. J Am Diet Assoc 1992;92:454-60. 19. Lukaski H. Methods for the assessment of human body composition: Traditional and new. Am J Clin Nutr 1987;46:537-56. 20. Behnke AR, Feen BG, Welham WC. Specific gravity of healthy men. JAMA 1942;118:495-8. 21. Colliver JA, Frank S, Frank A. Similarity of obesity indices in clinical studies of obese adults: A factor analytic study. Am J Clin Nutr 1983;38:640-7. 22. Keys A, Fidanza F, Karvonen MJ, Kimura N, Taylor HL. Indices of relative weight and obesity. J Chron Dis 1972;25:329-43. 23. Garn SM, Leonard WR, Hawthorne VM. Three limitations of the body mass index. Am J Clin Nutr 1986;44:996-7. 24. Van Itallie T. Health implications of overweight and obesity in the United States. Ann Intern Med 1985;103:983-8.
382
L i n d s a y et al
BMI and Percent Body Fat
A d d r e s s r e p r i n t r e q u e s t s to:
Carol A. Lindsay, MD Department of Obstetrics and Gynecology MetroHealth Medical Center 2500 MetroHealth Drive Cleveland, OH 44109
Received July 25, 1996. Received in revisedform October 28, 1996. Accepted November 21, 1996.
Copyright © 1997 by The American College of Obstetricians and Gynecologists. Published by Elsevier Science Inc
Obstetrics & Gynecology
Body mass index (BMI), defined as weight (kg)/height (m) 2, has been used as an estimate of obesity in both pregnant and nonpregnant individuals. Additionally, BMI has been used to predict various pregnancy outcomes and served as the basis for recommendations regarding appropriate weight gain and caloric intake during pregnancy. ~ Because BMI has been applied in both research and clinical settings as an indicator of obesity in pregnancy, 2'3 it is important to know how From the Department of Reproductive Biology, Case Western Reserve University at MetroHealth Medical Center, Cleveland, Ohio. Supported by NIH HD22965(PMC) and the General Clinical Research Center RR 00080.
VOL. 89, NO. 3, MARCH 1997
well this measure estimates percent body fat during pregnancy. The range for the correlation coefficient of BMI with percent body fat in nonpregnant w o m e n is between .60 and .82. 4 Based on our review of the literature using a MEDLINE search (1969-1996), the relationship between BMI and percent body fat has not been evaluated during pregnancy. Hence, this study was done to evaluate prospectively the longitudinal relationship between BMI and percent b o d y fat in nonpregnant w o m e n who subsequently conceived and to determine whether the relationship changed significantly over time. We hypothesized that BMI would be a poorer indicator of percent body fat in pregnancy than in the nonpregnant individual due to the relative increases in fat-free mass (such as plasma volume, placental mass, amniotic fluid (AF), and uterine mass) and that the strength of the correlation between percent body fat and BMI would decrease with advancing gestation.
Materials and Methods Our sample consisted of 27 healthy, multiparous, white, nonsmoking, nonobese w o m e n selected from a population of nonpregnant w o m e n with normal glucose tolerance who were contemplating pregnancy. Sixteen subjects participated in longitudinal studies 5 of glucose metabolism published previously. The enrollment period for this study was approximately 1 year. Pregravid studies were repeated every 6 months until the w o m e n conceived, with the measurements immediately preceding the pregnancies used as the baseline pregravid measurements. We recruited approximately 10-20% of subjects invited to participate in this study. Each subject had a pregravid BMI below 29.0. Mean maternal age at delivery was 32.3 years (range 23-40). Mean parity was 2.1 (range 1-3), and 89% of the subjects had at least a
0029-7844/97/$17.00
PI1 S0029-7844(96)00517-0
377
high school education. None of the women had medical problems, prior bowel surgery, or large variations in pre-pregnancy weight. Each subject had three estimates of body fat: pregravid, early gestation (defined as 12-17 weeks), and late gestation (defined as 33-36 weeks). Gestational age was determined by last menstrual period and confirmed in all subjects by first-trimester ultrasound. At each study period subjects were weighed on a calibrated scale and their heights were measured while the subjects stood with their backs against a wall. Body composition was estimated using underwater weighing or hydrodensitometry in the general clinical research center. Each subject was weighed underwater and the bathing suit weight was corrected/~ Residual lung volume was determined simultaneously by the nitrogen or helium washout technique. The subject breathed room air during the underwater weighing, followed by administration of 100% oxygen and measurement of the change in nitrogen or helium concentration of the expired air over 5-7 minutes. We used this value to calculate residual lung volume." Total body water increases by 0.7 L in early gestation, as d e m o n s t r a t e d by Lukaski et al. s The multicompartment model of body composition requires direct measurements of the various components of fatfree mass (bone mineral, non-bone mineral, water, and protein). 9 Theoretical equations derived to calculate body fat in pregnancy account for the increase in total body water, ~° but none have been tested in a prospective manner in early gestation. Therefore, in the current study the two-compartment model was used with equations derived in nonpregnant women before conception and in early gestation. Body fat was determined with the Keys and Brozek 11 equation (body fat = 420.1 divided by body density minus 381.3) for the pregravid and early gestation subjects. In late gestation there is an increase in total body water as measured with 2H21so.~2 During late gestation, body fat was estimated based on an equation derived at our institution by Catalano et al ~2(body fat = 518.57 divided by body density minus 476.30). This equation adjusts for increase in total body water by using a hydration constant of 0.76. j2 Percent body fat was calculated as body fat divided by total body weight. Once these data were collected, regression equations were created with BMI as the independent variable and percent body fat the dependent variable. Higher-order polynomial regression did not yield an improvement over simple linear regression; therefore, linear regression was used. The correlation between the two continuous variables was measured by Pearson's correlation
378 Lindsay et al
BM[ rout Percent Body Fat
Table 1. Body Composition at Each Period
Weight (kg) BMI (kg/m2) Body fat (%) Fat mass (kg) Fat-free mass (kg)
Pregravid
Early gestation
59.7 = 6.4 21.5 z 2.4 22.3 + 5.1 13.5 + 4.0 46.2 -+4.3
61.7 -+ 7.2 22.5 -+ 2.8 24.4 -+ 5.0 15.2 -+4.4 46.4 + 4.5
Late gestation 73.3 + 8.3* 26.0 ÷ 2.6* 26.6 -+ 4.8* 19.4 -+4.8 52.9 _+5.7
BMI = body mass index. Data are expressed as mean _+one standard deviation. *Changes over time significant by analysis of variance with repeated measures (P < .001). Changespregravid versus early gestation, pregravid versus late gestation, and early gestation versus late gestation also significant by Scheffetest, with an overall P < .05.
coefficient. In early gestation, the correlation between BMI and percent body fat, as determined by the Keys and Brozek ~1 equation, did not differ significantly from that determined by the equation derived by Catalano et al, ~2 the Siri 13 equation, or the equations by van Raaij et al ~° for 10 and 20 weeks' gestation; therefore, the Keys and Brozek 11 equation was used in the analysis of early gestation. The 95% confidence interval (CI) for predicting percent body fat was determined for the mean, upper, and lower extremes of BMI in each study period. Finally, an analysis of variance with repeated measures was done for each subject's change in weight, BMI, and percent body fat during pregnancy; and a secondary analysis using the Scheffe method of contrasts determined the difference in outcome measures at specific points. The value considered significant for this study was adjusted to P < .005 (Bonferroni's correction) due to the multiple comparisons made. The statistical software Statview 4.02 (Abacus Concepts, Berkeley, CA) was used for some statistical analyses.
Results Pregravid, early gestation, and late gestation weight; BMI; and percent body fat are listed in Table 1. Mean weight gain during early gestation was 1.96 kg, 87% of the increase in weight being fat. Late in gestation the percentage of weight gain due to the accumulation of fat decreased. Mean weight gain from pre-pregnancy until late gestation was 12.61 kg, and the percentage of the total weight gain in late gestation due to fat was only 47%. The correlation coefficient between pregravid BMI and percent body fat was r = 0.69 (Y = -9.491 + 1.478X, 95% CI 0.19, 0.76, P < .005). Given a mean pregravid BMI of 21.5, the predictive 95% CI for percent body fat is between 18.2 and 26.5%. The predictive 95% CI was also calculated at the extremes of BMI. For a BMI of
Obstetrics & Gynecology
18.1, the 95% predictive CI for percent body fat was between 12.8 and 21.7%. At the upper extreme, a BMI of 27.5 pregravid gave a predictive 95% CI between 26.0 and 36.2% body fat. The correlation between BMI and percent body fat in early pregnancy was r = 0.72 (Y = -11.762 + 1.63X, 95% CI 0.47, 0.86, P < .005). The difference between the correlation coefficient of BMI and percent body fat pregravid and in early gestation was not statistically significant by analysis of variance of repeated measures (P = .83). For a mean BMI of 22.3 in early gestation, the predictive 95% CI for percent body fat was between 20.0 and 29.0%. When measured in early gestation, a BM1 of 18.4 gave a 95% predictive CI for percent body fat between 13.5 and 23.1%. A BMI of 29.4 gave a 95% predictive CI for percent body fat between 30.5 and 41.7%. The correlation between BMI and percent body fat decreased in late pregnancy to r = 0.63 (Y = -3.735 + 1.166X, 95% CI 0.333, 0.817, P < .005). Neither the difference between the correlation coefficients of BMI and percent body fat pregravid and in late gestation nor the difference between the correlation coefficients in early and late gestation when compared by analysis of variance of repeated measures was statistically significant (P = .71 and .56, respectively). In late gestation, the mean BMI was 26.0 with a predictive 95% CI for percent body fat between 22.5 and 30.8°/,,. At the extremes, a BMI of 21.9 gave a 95°/,, predictive CI for percent body fat between 17.3 and 26.4%. At the u p p e r extreme, a BMI of 31.7 gave a 95% predictive CI for percent body fat between 28.3 and 38.1%. In all periods studied, the 95% CI for predicting percent body fat from BMI increased at the u p p e r extreme. The longitudinal changes in weight, BMI, and percent body fat with advancing gestation were all statistically significant (P < .001). All pair-wise comparisons of changes in weight, BMI, and percent body fat between pregravid and early and late gestation were statistically significant (P < .05). Finally, the ability to estimate percent body fat using weight alone was relatively poor compared with BMI, weight and percent body fat correlation coefficient r = 0.51 (P - .006), r = 0.54 (P .003), and r = 0.41 (P = .034) (pregravid, early gestation, and late gestation, respectively).
Discussion Our results indicate that the correlation between BMI and percent body fat in early pregnancy is consistent with that reported in the literature for nonpregnant subjects. However, although statistically significant, in early gestation BMI explains 52% of the variance in
VOL. 89, NO. 3, MARCH 1997
percent body fat and in late gestation it explains only 40% of the variance in percent body fat. Body composition changes during pregnancy. Of the 12.5-kg weight gain, approximately 3.5 kg is due to the accumulation of fat, the accretion of fat occurring predominately in the first half of pregnancy. In our sample, 29°/`, of the fat accumulation occurred during early gestation, whereas only 16% of the total weight gain occurred during this period. Products of conception represent 10, 20, 30, and 40% of the total weight gain at 10, 20, 30, and 40 weeks' gestation, respectively. ~4 This is consistent with our findings of decreasing amounts of total weight gain due to accretion of fat in late gestation. Percent body fat varies widely with the method used for its determination. Lederman et a115 found percent body fat to range from 21.5 as determined by total body water with dilution with D20, 22.3 with hydrodensitometry, and 29.1 by total body potassium in the same subject at 14 weeks. This is consistent with our findings of 24.4 percent body fat in early gestation as determined by hydrodensitometry. The same group also found a wide range in late gestation (defined as 37 weeks): 21.8 by total body water, 28.0 by hydrodensitometry, and 37.2 by total body potassium. Is Catalano et a112 also found a wide range in the third trimester: 34.4 by hydrodensitometry and 26.5 by total body water. Forsum et al 1~ used skinfolds to estimate percent body fat in two different groups of patients. In early gestation, percent body fat was 32.0 and 29.8, at 30 weeks it was 31.7 and 29.6, and at 36 weeks it was 30.2 and 28.3%. ~ In our study, percent body fat in late gestation as determined by hydrodensitometry was 26.6. The wide ranges of percent body fat are a reflection of violations of the many assumptions that must be made to calculate percent body fat with each different method in pregnant w o m e n as well as individual variations in the patients. Hydrodensitometry is one of the conllTlon research methods used in estimating body composition in humans and frequently is used as the standard by which other methods are evaluated. Hydrodensitometry, however, is an indirect measure of body composition. The main determinants of human body density are fat and bone. The density of bone is approximately 3.0, adipose tissue (which contains 83% fat, 15% water, and 2% protein) r'~s is 0.94, water is 0.996, and all other tissues are approximately 1.060. Fat appears to be the main factor affecting the specific density of an individual. The density of body fat is 0.900 g / m L , and all other body tissues, ie, fat-free mass, have a mean density of 1.100 g / m L ; therefore, low values of density indicate increasing obesity and greater values denote leanness. For the purposes of estimating body composition, the h u m a n body can be divided into components. Hydro-
Lindsay et at
BMI nnd Percent Body Fat
379
densitometry relies on a two-compartment model, which divides the body into fat and fat-free mass. Throughout gestation the amount of body water increases. In a longitudinal study of 15 women, Lukaski et al s found total body water to increase from a prepregnancy mean of 31.7 L to 32.2 L at 14-16 weeks and 38.8 L at 36-38 weeks. In a two-compartment model, an increase in total body water leads to a decrease in total body density. The density of water is approximately 0.996 and the density of fat-free mass is approximately 1.100; therefore, an increase in total body water leads to a decrease in density and an overestimation of body fat 12 when density equations are used for nonpregnant w o m e n during gestation. Keys and Brozek ~1 derived an equation for estimating body fat (420.1 divided by body density minus 381.3). Non-gravid b o d y density calculations assume a hydration constant between 0.72 and 0.73 of the free-fat mass. Although there is an increase in total body water in early gestation, no equations to estimate percent body fat from density have been tested in a prospective manner in early gestation. Because there is an increase in maternal total body water in late gestation, the use of equations with a hydration constant of 0.72-0.73 m a y result in overestimation of fat mass. Recently, an equation was derived for late gestation by Catalano and colleagues 12 (body fat = 518.57 divided by body density minus 476.30) that derived a hydration constant of 0.76 based on measurement of both density and total body water using 2H2180. Several assumptions apply with hydrodensitometry in estimating body fat. Hydrodensitometry as a measure of percent body fat assumes a constant hydration and a constant proportion of bone mineral to muscle in the fat-free body, 19 not varying by more than 0.004. 20 It also assumes that there is minimal air in the gut. It is not possible to measure air in the gut, and this causes an error of approximately 1.5% in the estimation of body fatJ 9 Finally, error also can be introduced from the measurement of residual lung volume, causing an error of _+0.003, which can be minimized with repeated determinations. These assumptions are used in nonpregnant w o m e n and we assume they apply to pregnant women as well, with the exception of the increase in total body water as noted previously. There is also a decrease in residual volume of approximately 200 mL in pregnancy, but the residual lung volume is measured at each density determination; therefore, it is assumed that this difference is accounted for in the density measurement. Although hydrodensitometry is probably the best method for determining the contribution of fat mass to weight in pregnancy, there are also potential technical
380 Lindsay et al
BMI and Percent Body Fat
difficulties in performing hydrodensitometry during pregnancy. Subjects must be highly motivated, cooperative, and able to remain submerged and motionless for the amount of time required to obtain the measurement. 17 The technique requires special equipment and about 1 hour to obtain the subject's density with residual lung volume measurement. In addition, the estimation of body composition by hydrodensitometry in pregnancy includes fetal mass, placental mass, and AF. The major methodologic disadvantage to the use of hydrodensitometry in pregnancy is that the estimation of percent body fat is dependent on the density of fat-free mass. Density of fat-free mass decreases in pregnancy due to an increase in total body water; therefore, fat mass will be overestimated using standard, nongravid equations. Hydrodensitometry is also expensive and time consuming and requires personnel skilled in the measurements and a cooperative patient. It is not practical for widespread application. Body mass index is used widely as an index of obesity. Obesity is defined as an increase in the percent of body fat I that has been estimated based on BMI by various organizations and investigators. A good measure of obesity should correlate highly with relative adiposity and be independent of height. 21 A number of body mass indices have been derived, including the weight-to-height ratio (W / H), Quetelet index (W / H 2), ponderal index (W1/3/H), Sheldon index (H/W1/3), Abde-Malek index (cWI2/H 3"3, where c = 4 × 106 for w o m e n and height is in centimeters), and percent desirable body weight expressed as percentage of the mean weight for a given height and sex. There was a strong intercorrelation between the various BMIs (ranging from 0.82 to 1.0). These indices are correlated strongly with weight and weakly with height. All BMIs correlated strongly with percent body fat as determined by densitometer (r = 0.87-0.92) in a selected nonpregnant female population. The correlation reported in the general population was not as strong (0.60-0.82). Quetelet index was found to be the most stable in young adults 22 and was used in the present study. The 95% C1 for the percent body fat as estimated from BMI was found to be wide, however, ranging from 10-12%. 4 This range was consistent with our findings in the current study. The 95% CI for predicting percent body fat from BMI was widest at the upper extreme of BMI both before gestation and during pregnancy. This would imply that BMI is less precise at predicting percent body fat at the u p p e r extremes of BMI. There are several pitfalls in the use of BMI as a measure of obesity. First, BMI is not completely independent of stature; the correlation coefficient between BMI and stature is approximately 0.12. Second, BMI is
Obstetrics & Gynecology
influenced nearly equally by the lean a n d fat compon e n t s of the body. 23 Two i n d i v i d u a l s can have the same BMI a n d still vary w i d e l y in the percent b o d y fat because muscle a n d fat contents are n o t necessarily correlated. 24 It has b e e n proposed that, on an i n d i v i d u a l basis, BMI not be used as a m e a s u r e of obesity, a l t h o u g h it m a y be useful in p o p u l a t i o n studies. 4 A simple m e a s u r e of obesity in p r e g n a n c y w o u l d be helpful because certain complications of p r e g n a n c y are associated w i t h i n c r e a s i n g obesity, i n c l u d i n g preeclampsia, gestational diabetes mellitus, h y p e r t e n s i o n a n d cardiovascular disease, s h o u l d e r dystocia, rate of cesarean delivery, w o u n d infections a n d dehiscence, anesthesia risks, increased blood loss, a n d other postp a r t u m complications. There is also controversy regarding the weight gain in p r e g n a n c y that results in the best outcome. A n objective m e a s u r e of percent b o d y fat m a y help further clarify i n d i v i d u a l caloric requirements. Excessive weight gain in p r e g n a n c y even in the absence of obesity is associated with both delivery of a large for gestational age infant a n d increased rate of cesarean delivery even after controlling for birth weight. Also, excessive weight gain is more likely to be retained after the p r e g n a n c y . ~ At least in n o n o b e s e w o m e n , one can use BMI to estimate percent b o d y fat, which m a y be a better predictor of adverse outcome than weight alone; however, further s t u d y in this area is necessary. The strength of this s t u d y is that it was conducted l o n g i t u d i n a l l y on the same subjects, decreasing the variability d u e to i n d i v i d u a l differences. Application of its results are limited, however, because it was cond u c t e d in healthy, m u l t i p a r o u s , n o n o b e s e w o m e n . Therefore, the results m a y not apply to a dissimilar group. Correlations m a y differ in obese patients or in patients with disease processes that m a y alter b o d y composition, such as excess e d e m a formation. Also, b o d y composition changes t h r o u g h o u t gestation; therefore, equations m u s t be developed specific for each period of gestation to d e t e r m i n e percent b o d y fat. O u r results indicate that BMI can be used as a m e a s u r e of percent b o d y fat in n o n o b e s e p r e g n a n t w o m e n , d e m o n s t r a t e d by a significant correlation as estimated by h y d r o d e n s i t o m e t r y that was consistent with that reported for the n o n p r e g n a n t subject. This correlation is superior to the correlation of weight alone with percent b o d y fat. Moreover, this correlation of BMI a n d percent b o d y fat is better in the pregravid a n d early gestation patient, w h e n m u c h of the a c c u m u l a t i o n of fat is occurring. However, in late gestation w h e n the maternal weight gain is affected more by growth of fetal tissues, maternal water, p l a s m a v o l u m e , a n d AF volume, this correlation decreases and should be used with greater caution.
VOL. 89, NO. 3, MARCH 1997
Body mass index is easy to calculate, can be determ i n e d cheaply, a n d relies on i n f o r m a t i o n generally obtained at a n y initial health assessment. For this reason, it w o u l d be r e c o m m e n d e d for general application to d e t e r m i n e percent b o d y fat in a p o p u l a t i o n ; however, d u e to the wide CI for predicting percent b o d y fat, BMI m a y not be precise e n o u g h for use in the i n d i v i d u a l patient. It is i m p o r t a n t to note that this s t u d y was c o n d u c t e d in a n o n o b e s e p o p u l a t i o n a n d the results m a y not be applicable to obese w o m e n .
Refel'enc¢s 1. Ramin SM, Cunningham FG. Obesity in pregnancy. In: Cunningham FG, MacDonald PC, Gant NF, LevenoKJ, Gillstrap LC 1II,eds. Williams obstetrics. 19th ed. Norwalk, Connecticut: Appleton & Lange, 1995:l 14. 2. Wolfe HM, Sokol RJ, Martier SM, Zador IE. Maternal obesity: A potential source of error in sonographic prenatal diagnosis. Obstet Gynecol 1990;76:339-42. 3. Wolfe HM, Zador IE, Gross TL, Martier SS, Sokol RJ. The clinical utility of maternal body mass index in pregnancy. Am J Obstet Gynecol 1991;164:1306-10. 4. Smalley KJ, Knerr AN, Kendrick ZV, Colliver JA, Owen OE. Reassessment of body mass indices. Am J Clin Nutr 1990;52:405-8. 5. Catalano PM, Tyzbir ED, Wolfe RR, Calles J, Roman NM, Amini SB, et al. Carbohydrate metabolism during pregnancy in control subjects and women with gestationaI diabetes. Am J Physiol 1993;364:E60-7. 6. Akers R, Buskirk ER. An underwater system utilizing "force cube" transducers. J AppI Physiol 1969;26:64932. 7. Darling C, Cournand A, Richards DW. Studies on the intrapuhnonary mixture of gases. An open circuit method for measuring residual air. J Clin Invest 1940;19:609-18. 8. Lukaski HC, Siders WA, Nielsen EJ, Hall CB. Total body water in pregnancy: Assessment by using bioelectrical impedance. Am J Clin Nutr 1994;59:578-85. 9. Lederman SA. Recent issues related to nutrition during pregnancy. J Arn Coil Nutr 1993;12:91-100. 10. van Raaij JM, Peek ME, Vermaat-Miedema SH, Schonk CM, Hautvast JG. New equations for estimating body fat mass in pregnancy from body density or total body water. Am J Clin Nutr 1988;48:24-9. 11. Keys A, Brozek J. Body fat in adult man. Physiol Rev 1953;33:245 325. 12. Catalano PM, Wong WW, Drago NM, Amini SB. Estimating body composition in late gestation: A new hydration constant for body density and total body water. Am J Physiol I995;268:E153-8. 13. Siri WE. Body compositionfrom fluid spaces and density: Analysis of methods. In: BrozekJ, Henschel A, eds. Techniques for measuring body composition. Washington, DC: National Academy of Science-NationalResearch Council, 1961:223-44. 14. Hytten F, Chamberlain G, Weight gain in pregnancy. In: Hytten F, Chamberlain G, eds. Clinical physiology in obstetrics. 2nd ed. Oxford: BlackwellScientific Publishers, 1991:173--203. 15. Lederman SA, Pierson RN, Wang J, Paxton A, Thornton J, Wendel J, et al. Body composition measurements during pregnancy. Bas Life Sci 1993;60:193-5. 16. Forsm'nE, Sadurskis A, Wager J. Estirnationof body fat in healthy Swedish women during pregnancy and lactation. Am J Clin Nutr 1989;50:465 73.
Lindsay et al
BMI at~d Percent Body Fat
381
17. Garrow JS. New approaches to body composition. Am J Clin Nutr 1982;35:1152-8. 18. Jensen MD. Research techniques for body composition assessment. J Am Diet Assoc 1992;92:454-60. 19. Lukaski H. Methods for the assessment of human body composition: Traditional and new. Am J Clin Nutr 1987;46:537-56. 20. Behnke AR, Feen BG, Welham WC. Specific gravity of healthy men. JAMA 1942;118:495-8. 21. Colliver JA, Frank S, Frank A. Similarity of obesity indices in clinical studies of obese adults: A factor analytic study. Am J Clin Nutr 1983;38:640-7. 22. Keys A, Fidanza F, Karvonen MJ, Kimura N, Taylor HL. Indices of relative weight and obesity. J Chron Dis 1972;25:329-43. 23. Garn SM, Leonard WR, Hawthorne VM. Three limitations of the body mass index. Am J Clin Nutr 1986;44:996-7. 24. Van Itallie T. Health implications of overweight and obesity in the United States. Ann Intern Med 1985;103:983-8.
382
L i n d s a y et al
BMI and Percent Body Fat
A d d r e s s r e p r i n t r e q u e s t s to:
Carol A. Lindsay, MD Department of Obstetrics and Gynecology MetroHealth Medical Center 2500 MetroHealth Drive Cleveland, OH 44109
Received July 25, 1996. Received in revisedform October 28, 1996. Accepted November 21, 1996.
Copyright © 1997 by The American College of Obstetricians and Gynecologists. Published by Elsevier Science Inc
Obstetrics & Gynecology