Scholten, Ralph R. MD; Oyen, Wim J. PhD, MD; Van der Vlugt, Maureen J. PhD, MD; Van Dijk, Arie P. J. PhD, MD; Hopman, Maria T. E. PhD, MD; Lotgering, Fred K. PhD, MD; Spaanderman, Marc E. A. PhD, MD
The association between low birth weight and increased blood pressure in adult life and mortality from cardiovascular disease has been well established.1,2 So far, possible mechanisms linking fetal growth restriction to higher blood pressures in adulthood may be found in persisting changes in vascular structure, including loss of elasticity in vessel walls3 or increased insulin resistance,4,5 altered glucocorticoid balance,6 or abnormal cholesterol metabolism.7
Low plasma volume in the absence of hypertension is related to the early, often latent, phase of chronic hypertension.8–10 At rest, two-thirds of the human blood volume is localized in the venous system.11 Quantitatively, most of the total venous return (47%) comes from veins within splanchnic organs and kidneys.11 Therefore, the splanchnic vascular bed is often regarded as the quantitatively most important system to “store” blood volume. Theoretically, a structural small venous abdominal compartment negatively affects the total plasma volume.
Asymmetrical fetal growth restriction is characterized by a disproportionately small abdominal circumference as a consequence of loss in liver, kidney, and intestinal volume.12 This is thought to originate from selective growth failure of splanchnic organs as a consequence of circulatory redistribution away from this abdominal region.13 Because only arteries have regenerative capacity,11 limited venous intrauterine development may have long-lasting effects on total plasma volume. In line with the fetal-origins-of-adult-disease theory, we hypothesized that women who were born small for gestational age (SGA) have lower plasma volume levels in adult life compared with those who were born appropriate or large for gestational age (LGA).
MATERIALS AND METHODS
We conducted a retrospective observational study in 347 consecutive healthy women with a history of hypertension in pregnancy in whom plasma volume was measured between January 2008 and December 2010. We used a database of women who were extensively evaluated in our tertiary hospital after hypertensive disorders in pregnancy. We recorded recalled birth weight and gestational age at birth. If women were unsure of these characteristics, they were excluded from this analysis. Women with pre-existing medical conditions or who used any medication were also excluded.
All measurements were performed, in the nonpregnant state, 6–12 months after a pregnancy complicated by preeclampsia, hemolysis, elevated liver enzymes, low platelets (HELLP) syndrome, or both. The diagnoses of gestational hypertension, preeclampsia, and HELLP syndrome were made according to the criteria of the American College of Obstetricians and Gynecologists.14 Except for their history of complicated pregnancy, all women were healthy.
Birth weight centiles of women included in this analysis were calculated according to the national reference population of women who were in the same era.15 The study group was stratified into four birth weight groups. Small for gestational age was defined as birth centile less than 10%. Women born appropriate for gestational age were subdivided in women who were 10th or more and less than 50th centile and women who were between 50th or more and 90th or less centile. Large for gestational age was defined as belonging to the more than 90% birth centile.
At the time of measurements, none of the women used hormonal contraceptives nor were they breast feeding. All women were white. The study was approved by the institutional review board of the Radboud University Nijmegen Medical Centre.
Plasma volume (mL) was measured using the 125I-human serum albumin indicator dilution technique. During the measurement, patients stayed in a semisupine position on a comfortable bed. An intravenous catheter was inserted in the left antecubital vein for the repetitive blood sampling. A standardized dose (0.2 MBq) of 125I-human serum albumin was injected intravenously in the right antecubital vein. Every 10 minutes a venous blood sample was taken from the contralateral intravenous catheter until 40 minutes after administration of the 125I-human serum albumin. Blood samples were analyzed using a gamma counter. Plasma volume was calculated by dividing the total injected radioactivity by the virtual volume-specific radioactivity at time zero, as described elsewhere.16 Plasma volume was indexed for body surface area (BSA, mL/m2). The technician measuring plasma volume was unaware of the birth weight and medical history of participants.
To explore possible confounding effects of hemodynamic, metabolic, endocrine, or lifestyle factors, we systematically determined these profiles in each participant on the same day as the plasma volume measurement. All participants collected urine in the 24 hours preceding the measurements. The 24-hour urine sample was assayed for sodium output (mmol/24 h), microalbumin corrected for creatinine output (g/mol creatinine), and creatinine (mmol/L).
Height and body mass were measured. Body surface area for the normalization of the plasma volume was calculated using the Dubois and Dubois formula.17 Subsequent measurements were performed after an acclimatization period of at least 20 minutes in the supine position in a temperature-controlled room (20–22°C). All measurements were performed in the morning in the fasting state (overnight).
Blood pressure and heart rate were measured oscillometrically at 3-minute intervals for 30 minutes at the right upper arm. We used the median values of nine consecutive measurements. Measurements were done with the cuff size recommended for the arm circumference. We recorded systolic and diastolic blood pressures (mm Hg) and heart rate (beats per minute). Pulse pressure (mm Hg) was calculated as the difference between median systolic and diastolic blood pressures. We calculated mean arterial pressure (mm Hg) using the formula: (2×diastolic blood pressure+systolic blood pressure)/3.
Echocardiographic measurements were obtained by an experienced cardiology technician Measurements were performed in the left lateral position using a cross-sectional phased-array echocardiographic Doppler system. Left atrial diameter was measured in the four-chamber view. Subsequently left ventricular outflow trajectory velocity integral and left ventricular outflow diameter were measured. Heart rate was determined by the reciprocal of the relative risk interval of the electrocardiogram measured during the echo Doppler measurements. This heart rate was used only for the calculation of cardiac output. Stroke volume was calculated by multiplying the left ventricular outflow tract velocity integral and the left ventricular outflow tract diameter. Cardiac output was calculated as stroke volume×heart rate. Total peripheral vascular resistance was calculated as eighty times the mean arterial pressure divided by the cardiac output (80×mean arterial pressure/cardiac output). Global vascular compliance was calculated from stroke volume and pulse pressure using the equation: global vascular compliance=stroke volume/pulse pressure (mL/mm Hg).
Venous blood samples were taken from the in situ intravenous catheter and analyzed for metabolic parameters: glucose (mmol/L), insulin (milliunits/L), total cholesterol, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, triglycerides (mmol/L), sex hormone levels (progesterone (nmol/L), and estradiol (pmol/L) and renal function (serum creatinine level (micromole/L). Homeostasis model of assessment–insulin resistance was calculated using the formula: [fasting serum insulin (milliunits/L)×fasting plasma glucose (mmol/L)/22.5]. Creatinine clearance was calculated from the 24-hour urine creatinine concentration and the plasma creatinine concentration corrected for BSA and expressed in mL/min/1.73 m2.
Daily activity was assessed with a validated questionnaire (Short Questionaire to Assess Health-enhancing physical activity).18 Participants were asked to describe an average week. Intensity was expressed as metabolic equivalent task (MET), a physiologic concept of expressing the energy of physical activities as multiples of resting metabolic rate, set by convention to 3.5 mL O2×kg−1×min−1. Activities were subdivided into three intensity categories 2.0 to less than 4.0 MET (light), 4.0 to less than 6.5 MET (moderate), and 6.5 MET or greater (vigorous) using Ainsworth's compendium of physical activities.19 Activities below 2 MET value were not included, because they are considered to contribute negligibly to habitual activity levels. The Short Questionaire to Assess Health-enhancing physical activity quantifies commuting and work-related activities, leisure activities (including sports), and household activities. Physical activity is presented as minutes per day (median [interquartile range]) for each intensity category (light, moderate, and vigorous). For each individual, the sum of the physical activity in each intensity category (total daily activity level; minutes per day) was used in the multivariable analysis.
Data were expressed as means±standard deviation or as median with interquartile range for normally and nonnormally distributed data, respectively. Normality of each variable was evaluated using Kolmogorov-Smirnov tests. Trends were analyzed by linear regression of the adult characteristics on birth centile as a continuous variable and not categorical.
We calculated the change in adult plasma volume (mL) associated with a 10 centile increase in birth centile by linear regression analysis. The effects of possible confounding variables based on their known potential association with adult plasma volume included: current age, body mass index (calculated as weight (kg)/[height (m)]2), smoking, daily activity level, progesterone level, estradiol level, mean arterial pressure, global vascular compliance, insulin level, glomerular filtration rate, 24-hour sodium output, and interval between delivery and measurement. In the end, adjustments were made by multiple linear regression for those factors that also related to birth centile (BSA, mean arterial pressure, total vascular resistance, creatinine clearance and total 24-hour sodium output).
To identify factors contributing to the adult plasma volume status, we first tested univariably (Pearson's) which variables correlated with total adult plasma volume (mL). Subsequently, factors correlating with plasma volume (P<.05) were introduced in a backward stepwise linear regression analysis to weigh the potential determinants of the adult total plasma volume (mL). To quantify the independent contribution of each variable in our final model for adult plasma volume, we consecutively removed factors from our model starting with the quantitatively most important factor based on the largest change in the explained variance (R2) after removal. We present the explained variance (R2) of the final model and the R2 of the subsequent models after consecutive removal of a variable. Statistical significance (two-sided P value) was set at ≤.05. The statistical analyses were performed using the standard statistical software package SPSS 16.0.
Two hundred eighty women were available for analysis after exclusion of 67 women. Exclusion followed after inability to recall their own birth characteristics (n=51); 14 women had to be excluded because of pre-existing medical conditions or use of medication. Two women in whom plasma volume could not be adequately determined as a result of technical reasons were also excluded (Fig. 1). Women who were excluded for not being able to recall their own birth weight (n=51) did not differ from the women included in our study (n=280) in plasma volume (2,775 compared with 2,761 mL, P=.58), current BSA (1.81 compared with 1.81 m2, P=.71), mean arterial pressure (85 compared with 84 mm Hg, P=.62), total vascular resistance (1,297 compared with 1,329 dynes×s/cm5, P=.42), glomerular filtration rate (109 compared with 110 mL/min/1.73 m2, P=.81) and total 24-hour sodium output (134 compared with 135 mmol/24 hour, P=.75).
Table 1 demonstrates the general adult characteristics and the characteristics at birth of the women included in the analysis stratified by the four birth centile groups. With increasing birth weight centile, both adult height and weight increased; this pattern in body composition was also reflected in the BSA (P for trend=.004). The gestational age at birth did not differ between birth centile groups.
Linear regression analysis demonstrated that each 10 centile change in birth weight relates to an associated with an average change of 46.6 mL (95% confidence interval 30.8–62.3) in adult plasma volume (Fig. 2). When plasma volume was normalized for adult BSA, the positive correlation between birth centile and adult plasma volume persisted (P for trend<.001) (Table 2). Women born SGA had 11% lower plasma volume (mean, 1,349±200 mL/m2) compared with women who were appropriate for gestational age 10th or more and less than 50th centile (mean, 1,518±156 mL/m2). Furthermore, women who were LGA had 10% higher plasma volume (mean, 1,663±229 mL/m2) compared with women who were appropriate for gestational age 10th or more and less than 50th centile. Plasma volume in women who were appropriate for gestational age between 50th or greater and 90th or less centile (mean, 1,544±137 mL/m2) did not differ from appropriate for gestational age 10th or more and less than 50th centile.
Hemodynamically, birth centile related inversely to the mean arterial blood pressure (P for trend=.04) (Table 2). This trend was reflected in both systolic and diastolic blood pressures. Total vascular resistance in adulthood tended to decrease with increasing birth centile (P for trend=.09). We found no differences in cardiac performance between birth centile groups. Also, physical activity levels did not differ between birth centile groups.
Insulin level, homeostasis model of assessment–insulin resistance and triglyceride level were higher in women who were SGA compared with women who were appropriate for gestational age 10th or more and less than 50th centile (Table 3). Metabolically we found no linear trends between birth centile and glucose metabolism or lipid concentrations. In contrast, glomerular filtration rate increased with birth weight centile (P for trend=.05). The 24-hour sodium output tended to increase with birth centile (P for trend=.08). When the specific cardiovascular risk factors were clustered according to the World Health Organization criteria for the metabolic syndrome, women who were SGA had the highest prevalence of metabolic syndrome (23%). Estrogen and progesterone levels did not differ between groups.
To explore possible confounding effects of the recent pregnancy, Table 4 summarizes the recent obstetric history of the included women. By inclusion, all women met the criteria for pregnancy-induced hypertension in previous pregnancy. In our study population, 176 (63%) women additionally met the criteria for preeclampsia. As indicated in Table 4, the obstetric outcomes did not differ between birth centile groups. Also, time interval between delivery and plasma volume measurements was the same in all birth centile groups.
When the relation between birth centile and total adult plasma volume (mL) was adjusted for current BSA, mean arterial pressure, total vascular resistance, glomerular filtration rate, and a total 24 hours of sodium output, each 10 centile change in birth weight is associated with an average change of 32.1 mL (95% confidence interval 19.6–44.6) in adult plasma volume.
To identify predictors of adult plasma volume, we firstly explored univariably which factors correlated with adult plasma volume (mL). Plasma volume correlated with BMI (r=0.44; P<.01), birth weight centile (r=0.36; P<.01), glomerular filtration rate (r=0.29; P<.01), total vascular resistance (r=−0.23; P<.01), daily activity level (r=0.23; P<.01), global vascular compliance (r=0.23; P<.01), 24-hour sodium output (r=0.21; P=.01), and mean arterial pressure (r=−0.13; P<.01). Subsequent multivariable backward regression analysis revealed that only BMI, birth weight centile, daily activity level, and glomerular filtration rate contributed independently to the adult plasma volume (R2 final model=0.48). Quantitatively, BMI was the most important determinant in our model for total adult plasma volume, as indicated by the largest drop in explained variance (−0.17) after removal of the factor BMI from our final model (Table 5). Subsequent removal of birth centile indicates that the birth centile contributes for an additional 14% of the explained variance in total adult plasma volume. Similarly, daily exercise level and glomerular filtration rate accounted for 10% and 7% of the variance, respectively. Fifty-two percent of the variance in total plasma volume remains unexplained by our model.
Our study demonstrates that women with a history of hypertension in pregnancy and who were born SGA are predisposed for lower plasma volumes in adult life compared with counterparts who were born appropriate or LGA. Low plasma volume is a condition known to precede chronic hypertension.9,20 The association between birth weight centile and subsequent adult plasma volume is independent of current body composition, smoking, blood pressure, physical activity level, cardiac performance, glucose metabolism, lipid levels, kidney function, and levels of sex hormones.
Formerly preeclamptic women have an increased risk of developing cardiovascular disease, suggesting a common risk profile.21 Women who had gestational hypertension have a similar, but lower, risk of future hypertension and cardiovascular disease as those who had preeclampsia.21 Previous studies also demonstrated associations between low plasma volume and preeclamptic pregnancies.16,22,23 Etiology of this low plasma volume was largely unknown. This study identifies four independent determinants of the total adult plasma volume: body composition (BMI), birth weight centile, daily exercise level, and the glomerular filtration rate.
Reduced plasma volume could be an interesting factor linking, at least partially, intrauterine development with hypertension in adulthood and remote cardiovascular disease. One process associated with fetal disproportionate growth is the redistribution of blood flow in favor of the brain, heart, and adrenal glands, but away from other organs, including splanchnic organs and the kidneys.12,24,25 Interestingly, the size of splanchnic organs and kidneys is reduced in children born SGA.12 Our data suggest that reduced size of the splanchnic vascular bed may have a long-lasting effect on adult plasma volume status.
Kidneys of individuals born SGA are known to contain fewer nephrons, a condition that is thought to initiate hypertension.26,27 Nephrogenesis is completed relatively late in gestation by 34–36 weeks, increasing the effect of fetal compromise on renal development.26,27 People with fewer nephrons are likely to have a relatively high glomerular filtration rate in each available nephron. Over time, hyperfiltration may cause glomerular injury and ultimately glomerular loss and the development of hypertension.26
In our study, the glomerular filtration rate correlated with adult plasma volume. However, the relationship between birth weight centile and subsequent adult plasma volume was independent of glomerular filtration rate. This does not rule out single nephron hyperfiltration.
Our data concerning hemodynamic and metabolic variables are consistent with earlier studies.4,7,28 We confirm an inverse relationship between blood pressure and birth weight centile. Demonstrating this association in a young (younger than 40 years) female population that is considered to be healthy is remarkable. Our data indicate that women who were born SGA and who have a history of hypertension in pregnancy have lower cardiovascular reserve capacity (ie, lower plasma volume) at comparable arterial demands (ie, cardiac index) when compared with women who were normal or LGA. Insulin levels were evidently higher in women who were SGA. We observed that women who were SGA and with a history of hypertension in pregnancy have an almost 23% risk of metabolic syndrome compared with 10% in women who were appropriate for gestational age with a similar obstetric history.
Our data are applicable to selected population of women with a history of hypertensive complications in pregnancy. This could potentially amplify certain associations between birth weight centile and current metabolic and hemodynamic variables. However, this does not per se limit generalization of our results to the common female population of comparable age without a history of hypertension in pregnancy, because a recent cohort study of formerly preeclamptic women indicates that at least hemodynamic variables resolve largely within 6 months after giving birth, although further recovery may be possible up to 2 years postpartum.29 Furthermore, differences in recent obstetric outcome did not affect the relation between birth centile and adult plasma volume.
In this study, we used self-reported birth weight that could have resulted in a recall bias. Nonetheless, numerous studies in women with a comparable age (30–40 years) demonstrated a high accuracy of self-reported of birth weight (Spearman's correlation with state birth records varied between: r=0.74 and r=0.85).30,31 Therefore, we feel that recall bias has not substantially affected our observations. Because we did not observe differences between those who were able to recall their own birth weight and those who were not, we do not expect exclusion of this latter group affected our results. This study was unable to explore all potential mechanisms explaining the relationship between birth centile and subsequent adult plasma volume. We have no data directly evaluating the sympathetic system, the renin–aldosterone–angiotensin pathway, or potential effects of the glucocorticoid axis. Effects of these systems are considered to relate inversely to the plasma volume.32 Because we excluded women using medication against chronic hypertension, a condition likely to accompany alterations in these axes, we assume no substantial bias from these systems on our findings. Nevertheless, we were unable to quantify their contribution to the adult plasma volume status in this study. Unfortunately, we had no data concerning abdominal circumference at birth, which might have reflected size of the splanchnic organs even better.
A particular strength of this the study is the measurement of plasma volume in a large cohort of women in combination with the many important covariates that might determine adult plasma volume. Resultingly, we were able to analyze, exclude, or both of the potential confounding effects of these covariates and eventually weigh the contribution of predictors of adult plasma volume.
In conclusion, birth weight centile relates linearly to adult plasma volume in women with a history of hypertension in pregnancy. Future studies will have to elucidate the effects of interventions aimed at modifying plasma volume on the recurrence of hypertension in pregnancy and future cardiovascular disease.
1. Barker DJ, Osmond C, Golding J, Kuh D, Wadsworth ME. Growth in utero, blood pressure in childhood and adult life, and mortality from cardiovascular disease. BMJ 1989;298:564–7.
2. Osmond C, Barker DJ, Winter PD, Fall CH, Simmonds SJ. Early growth and death from cardiovascular disease in women. BMJ 1993;307:1519–24.
3. Martyn CN, Barker DJ, Jespersen S, Greenwald S, Osmond C, Berry C. Growth in utero, adult blood pressure, and arterial compliance. Br Heart J 1995;73:116–21.
4. Barker DJ, Hales CN, Fall CH, Osmond C, Phipps K, Clark PM. Type 2 (non-insulin-dependent) diabetes mellitus, hypertension and hyperlipidaemia (syndrome X): relation to reduced fetal growth. Diabetologia 1993;36:62–7.
5. Phillips DI, Barker DJ, Hales CN, Hirst S, Osmond C. Thinness at birth and insulin resistance in adult life. Diabetologia 1994;37:150–4.
6. Edwards CR, Benediktsson R, Lindsay RS, Seckl JR. Dysfunction of placental glucocorticoid barrier: link between fetal environment and adult hypertension? Lancet 1993;341:355–7.
7. Barker DJ, Martyn CN, Osmond C, Hales CN, Fall CH. Growth in utero and serum cholesterol concentrations in adult life. BMJ 1993;307:1524–7.
8. Cohn JN. Relationship of plasma volume changes to resistance and capacitance vessel effects of sympathomimetic amines and angiotensin in man. Clin Sci 1966;30:267–8.
9. Julius S, Pascual AV, Reilly K, London R. Abnormalities of plasma volume in borderline hypertension. Arch Intern Med 1971;127:116–9.
10. Safar ME, Weiss YA, Levenson JA, London GM, Milliez PL. Hemodynamic study of 85 patients with borderline hypertension. Am J Cardiol 1973;31:315–9.
11. Guyton AC. The venous system and its role in the circulation. Mod Concepts Cardiovasc Dis 1958;27:483–7.
12. Latini G, De MB, Del VA, Chitano G, De FC, Zetterstrom R. Foetal growth of kidneys, liver and spleen in intrauterine growth restriction: ‘programming’ causing ‘metabolic syndrome’ in adult age. Acta Paediatr 2004;93:1635–9.
13. Nathanielsz PW, Hanson MA. The fetal dilemma: spare the brain and spoil the liver. J Physiol 2003;548:333.
14. Kloosterman GJ. Intrauterine growth and intrauterine growth curves [in Dutch]. Ned Tijdschr Verloskd Gynaecol 1969;69:349–65.
15. Diagnosis and management of preeclampsia and eclampsia. ACOG Practice Bulletin No. 33. American College of Obstetricians and Gynecologists. Obstet Gynecol 2002;99:159–67.
16. Spaanderman M, Ekhart T, van Eyck J, de Leeuw P, Peeters L. Preeclampsia and maladaptation to pregnancy: a role for atrial natriuretic peptide? Kidney Int 2001;60:1397–406.
17. Du Bois D, Du Bois EF. A formula to estimate the approximate surface area if height and weight be known. Nutrition 1989;5:303–11.
18. Wendel-Vos GC, Schuit AJ, Saris WH, Kromhout D. Reproducibility and relative validity of the short questionnaire to assess health-enhancing physical activity. J Clin Epidemiol 2003;56:1163–9.
19. Ainsworth BE, Haskell WL, Leon AS, Jacobs DR Jr, Montoye HJ, Sallis JF, et al.. Compendium of physical activities: classification of energy costs of human physical activities. Med Sci Sports Exerc 1993;25:71–80.
20. Lebel M, Grose JH, Blais R. Alteration of the volume–renin component in borderline essential hypertension. Clin Exp Hypertens 1981;3:39–54.
21. Bellamy L, Casas JP, Hingorani AD, Williams DJ. Pre-eclampsia and risk of cardiovascular disease and cancer in later life: systematic review and meta-analysis. BMJ 2007;335:974.
22. van Beek, Ekhart TH, Schiffers PM, van EJ, Peeters LL, de Leeuw PW. Persistent abnormalities in plasma volume and renal hemodynamics in patients with a history of preeclampsia. Am J Obstet Gynecol 1998;179:690–6.
23. Aardenburg R, Spaanderman ME, van Eijndhoven HW, de Leeuw PW, Peeters LL. A low plasma volume in formerly preeclamptic women predisposes to the recurrence of hypertensive complications in the next pregnancy. J Soc Gynecol Investig 2006;13:598–603.
24. Jensen A, Garnier Y, Berger R. Dynamics of fetal circulatory responses to hypoxia and asphyxia. Eur J Obstet Gynecol Reprod Biol 1999;84:155–72.
25. Rudolph AM. The fetal circulation and its response to stress. J Dev Physiol 1984;6:11–9.
26. Brenner BM, Chertow GM. Congenital oligonephropathy: an inborn cause of adult hypertension and progressive renal injury? Curr Opin Nephrol Hypertens 1993;2:691–5.
27. Mackenzie HS, Lawler EV, Brenner BM. Congenital oligonephropathy: the fetal flaw in essential hypertension? Kidney Int Suppl 1996;55:S30–4.
28. Morgan AR, Thompson JM, Murphy R, Black PN, Lam WJ, Ferguson LR, et al.. Obesity and diabetes genes are associated with being born small for gestational age: results from the Auckland Birthweight Collaborative study. BMC Med Genet 2010;11:125.
29. Berks D, Steegers EA, Molas M, Visser W. Resolution of hypertension and proteinuria after preeclampsia. Obstet Gynecol 2009;114:1307–14.
30. Troy LM, Michels KB, Hunter DJ, Spiegelman D, Manson JE, Colditz GA, et al.. Self-reported birthweight and history of having been breastfed among younger women: an assessment of validity. Int J Epidemiol 1996;25:122–7.
31. Wodskou PM, Hundrup YA, Obel EB, Jorgensen T. Validity of self-reported birthweight among middle-aged and elderly women in the Danish Nurse Cohort Study. Acta Obstet Gynecol Scand 2010;89:1134–9.
32. Convertino VA, Keil LC, Bernauer EM, Greenleaf JE. Plasma volume, osmolality, vasopressin, and renin activity during graded exercise in man. J Appl Physiol 1981;50:123–8.