Mean Red Cell Volume as a Correlate of Blood Pressure
Background Clinical studies suggest that hypertensives have lower mean corpuscular volume (MCVs) than do normotensives. Epidemiological studies show no relation or higher MCVs. In the present study of elderly men (71 to 93 years of age) of the Honolulu Heart Program, elements of both findings are confirmed.
Methods and Results Three groups are identified: (1) those receiving no hypertension treatment, (2) those receiving treatment with any diuretic, and (3) those receiving treatment with nondiuretics only. MCV is lower in group 3 than in group 1 (−0.85 fL, P<.001) but the same in groups 1 and 2. Within groups 1 and 3, inverse relations of −0.22 and −0.09 mm Hg/fL (P<.05) are noted for systolic (SBP) and diastolic (DBP) blood pressures. No relations are observed in group 2. MCV and red blood cell count (RBC) are inversely correlated (r=−.45). In group 2, adjustment for RBC unmasks a direct relation between MCV and SBP (0.5 mm Hg/fL, P=.02) and DBP (0.3 mm Hg/fL, P=.02). In groups 1 and 3, relations between SBP and MCV are lost after adjustment for RBC (0.005 mm Hg/fL). For DBP, adding RBC plus an MCV×RBC interaction is significant (P<.001). DBP is 5 mm Hg greater in the highest RBC quartile than in the lowest. A +3 mm Hg difference between extreme MCV quartiles is noted only at high RBC levels.
Conclusions The relation between blood pressure and red cell measures is probably mediated by whole blood viscosity. Hematocrit is a determinant of whole blood viscosity. Viscosity affects peripheral resistance to blood flow, and peripheral resistance affects DBP. At high RBC levels, MCV may be “downregulated.” This may lower whole blood viscosity and partially reduce DBP without compromising flow.
Recent reports suggest that people with hypertension have lower MCVs than do subjects with normal blood pressure.1 2 Two animal models support such a relation, with derangement in transmembrane sodium flux control implicated as a cause of cell volume regulation and of blood pressure control, particularly DBP.3 However, a recent cross-sectional population-based study of 317 employed men in Italy with a low prevalence of elevated blood pressure demonstrated no relation between MCV and blood pressure, with or without adjustment for alcohol and smoking behaviors.3 Similarly, preliminary analysis of data from the Gubbio Population Study in Italy indicates that men and women diagnosed as hypertensive had statistically significantly higher MCVs than did normotensive control subjects.4
There is significant literature documenting relations of blood pressure with the erythrocyte measures of RBC, hematocrit, and hemoglobin in clinical and epidemiological studies but not with MCV.
Since the turn of this century,5 clinical studies have indicated a class of patient with positively correlated erythrocyte measures (eg, RBC, hematocrit, and hemoglobin) and blood pressure. This condition, variously called “stress polycythemia,” “relative polycythemia,” “spurious polycythemia,” and “pseudopolycythemia,” has been used to describe a clinical condition in which a raised hematocrit is associated with a normal total-body red cell mass, with normal or slightly reduced total-body plasma volume, and with hypertension, obesity, “stress,” and heavy cigarette smoking.6 The topic was reviewed in 1987 by Isbister,7 and a significant role for whole blood viscosity is identified as a mediator of these relations.
Population-based studies extending to the mid-1960s uniformly demonstrate positive relations of blood pressure with the erythrocyte measures of RBC, hematocrit, and hemoglobin.8 9 10 11 12 13 14 15 16 Almost every one of these studies demonstrate positive correlation coefficients on the order of .1 to .2 with DBP, but weaker positive relations involving SBP are noted only in some subsets of subjects.
In neither the clinical studies5 6 7 nor the epidemiological studies8 9 10 11 12 13 14 15 16 was the role of MCV in relations with blood pressure examined, except for the four studies reporting apparently conflicting results.1 2 3 4
In seemingly healthy populations, the MCV appears to be negatively correlated with RBC consistently with a coefficient of −.4 to −.5.3 17 The basis for this negative relation is not clear. Because hematocrit is directly proportional to the simple product of RBC and MCV, these statistical and mathematical relations produce circumstances in which negative confounding may operate to obscure the nature of relations between blood pressure and specific erythrocyte measures, particularly MCV.
The fourth examination of men of the Honolulu Heart Program offers the opportunity to assess whether the MCV is directly or inversely related to blood pressure. This assessment demonstrates that elements of both claims are true, depending on whether one is examining differences in average MCV between groups of nontreated men and of men treated for hypertension or relations within groups, focusing on the regression relation between blood pressure and MCV. The results of the latter relations, in particular, suggest a complex interrelation between DBP and erythrocyte indexes that is hypothesized to reflect alterations in long-term pressure/flow regulation mediated by blood viscosity mechanisms.
The Honolulu Heart Program is a population-based longitudinal study initiated in 1965 to identify risk factors for cardiovascular disease in Japanese-American men living in Hawaii. A baseline examination of 8006 men who were living on the island of Oahu and were born between 1900 and 1919 was conducted between 1965 and 1968. The men’s ages ranged between 45 and 68 years at the time of their examination. Subjects were identified through Selective Service records, and the results of recruitment and the methods of data collection have been described previously.18 19
Of these 8006 men, 4678 were eligible for participation in examination 4 between 1991 and 1993. The examination and informed consent protocol were reviewed and approved by the Kuakini Medical Center Research Institutional Review Board, and all procedures were in accordance with institutional guidelines. A total of 3741 examinations were conducted among survivors, including 3194 clinic visits, 485 home visits, and 62 nursing home visits. In addition to the 3741 examinations, a long telephone interview was obtained for 104 men, and a short telephone interview was obtained for 718 men. Thus, some information was available for a total of 4563 men. Of the 3741 men participating in some form of examination, 3363 of them had complete measurements on all covariates of interest in this study, including red cell measurements.
Baseline Interview and Clinical Examination
Participants were asked to fast for 12 hours before the clinic visit. Procedures included seated, supine, and standing blood pressure measures and resting 12-lead ECG followed by venipuncture and a physical examination. After the fasting blood sample was drawn, other examination components were scheduled throughout the morning. Anthropometric measurements included standing height without shoes and weight.
Red cell indexes were measured on blood collected in a standard 5-mL EDTA specimen tube with a Coulter Counter S Plus (Coulter Corp). The tube was stored at room temperature until laboratory analysis 2 to 4 hours later. Specimens from home visits were analyzed 4 to 6 hours after specimen collection. The Coulter Counter directly measures RBC and MCV and calculates hematocrit as a scaled product of the two.
Blood pressure measurements were obtained on five occasions during the examination; these included two sitting measurements, a supine measurement, and a standing measurement with a conventional mercury sphygmomanometer and a sitting measurement with a Hawksley random zero mercury sphygmomanometer. Technicians were trained to use standardized blood pressure measurement techniques, which included measurement of midarm circumference and selection of appropriate-sized cuffs, stethoscope placement, determination of peak inflation level, appropriate deflation rate (2 mm Hg/s), identification of first-phase (systolic) and fifth-phase (diastolic) Korotkoff sounds, avoidance of digit preference, and proper operation of the Hawksley random zero sphygmomanometer. Technicians were initially certified, recertified 1 month later, and thereafter recertified every 6 months. An average of the five measurements was used as the measure of blood pressure after demonstrating that similar patterns of relation existed for separate measures.
Subjects were asked to bring to the examination all prescription medications, which were recorded by name. Prescription medications were being taken by 2783 men, of whom 1553 were taking antihypertensive medications. A direct question about the use of prescribed antihypertensive agents identified 333 men who denied taking such drugs but who nevertheless reported taking a drug that could be prescribed for hypertension control. Although a portion of these men may have been prescribed such medications for other conditions, such as congestive heart failure, angina, and/or arrhythmias, they were identified as taking antihypertensive agents in the belief that many of them were probably unaware of why they were taking a medication or that the impact of such medication use on blood pressure/erythrocyte measure relations would be unrelated to indication for treatment.
With the use of the strategy of Cirillo et al,20 men with complete data on all covariates were classified into three groups based on treatment status and the known impact of diuretics on plasma volume, which affects red cell measurements.21 These groups were (1) 1810 men taking no antihypertensive drugs, including men having elevated blood pressure of >140 mm Hg SBP or >90 mm Hg DBP; (2) 477 men taking diuretics either alone or in combination with other types of antihypertensive agents; and (3) 1076 men taking nondiuretic antihypertensive agents. Nondiuretic agents included vasodilators, angiotensin-converting enzyme inhibitors, calcium channel blockers, β-blockers, and combinations of these.
Analyses were conducted with the sas program, version 6.08, for Windows-based desktop personal computers.22 Descriptive statistics such as mean and SD values among the medication groups were calculated and compared with the use of proc t test, and multiple linear regression modeling was done with the use of proc reg. The progression of analyses is detailed in “Results.” Significance for two-sided type I statistical error was arbitrarily chosen to be α<.05.
A series of three models was used to sequentially examine the relation between SBP or DBP and (1) MCV; (2) MCV plus RBC; and (3) MCV, RBC, and their interaction (MCV×RBC). The interaction of MCV and RBC is directly proportional to the hematocrit (hematocrit ∝MCV×RBC), and, indeed, the Coulter Counter determines the hematocrit by directly measuring the MCV and the RBC and then calculating the scaled product of the two. Thus, models containing the terms MCV, RBC, and hematocrit are equivalent to a model testing the significance of the statistical interaction between MCV and RBC. For this reason, although RBC and hematocrit are highly correlated (r≈.9), because they are mathematically related as a multiplicative interaction, issues of colinearity are not relevant.
Reported relations between red cell measures and blood pressure were statistically adjusted for age, weight, and height. In analyses where men were pooled from two treatment groups, group membership was coded as a two-level covariate and taken into account as a statistical adjustment factor. Additional analyses that were also adjusted for smoking and alcohol consumption habits had no impact in altering the magnitude or significance of regression coefficients or of differences in mean values of blood pressure among categories of red cell measures. Thus, only age-, weight-, height-, and, as appropriate, treatment group–adjusted statistics are reported.
Men receiving treatment for hypertension are notably heavier (2.5 to 3.0 kg) than men not under treatment (Table 1⇓). Men exclusively receiving nondiuretics are, on average, ≈7 to 9 months younger than the no-treatment and diuretic-treatment groups. SBP is notably higher in treated groups—3.6 mm Hg higher in the diuretic group and almost 10 mm Hg higher in the group treated only with nondiuretics.
RBC and hematocrit are significantly higher in both treatment groups. However, MCV is notably lower in men treated exclusively with nondiuretic agents, whereas no difference from the no-treatment group is noted for men treated with any diuretics. These differences, although clinically small, are 16% to 19% of respective SDs.
The regression relation predicting SBP and DBP from MCV changed dramatically as RBC and then hematocrit (∝MCV×RBC) are sequentially added to age-, weight-, and height-adjusted models (Table 2⇓).
For SBP, the regression coefficient for MCV is initially negative for the no-treatment group (1) and nondiuretic-treated group (3) and statistically significant (P=.007) for the former group. However, with the addition of RBC as a covariate, the MCV regression coefficients of both groups are markedly decreased in magnitude—effectively to zero—whereas the regression coefficients for RBC are strongly significant and approximately of the same positive magnitude in the two groups. The addition of hematocrit, which is effectively a model for statistical interaction, accounts for no significant additional variation in SBP, as indicated by the lack of statistical significance of the hematocrit regression coefficient.
For the group treated with any diuretic (group 2), the regression coefficient for MCV predicting SBP is slightly positive although not statistically significant (P=.43). However, with the addition of RBC as a covariate, the magnitude of the MCV regression coefficient increases more than threefold and becomes significant (P=.02). The general pattern of SBP among joint categories of RBC and MCV confirms the elevation of blood pressure with increased count but what may be a slight flattening of the direct relation between blood pressure with MCV at higher RBC counts (Fig 1⇓). This suggested pattern of relation involving MCV is in contrast to the other two treatment groups (discussed later), although it may be a random effect given the marginal statistics when hematocrit is added to the model. The addition of hematocrit, effectively a test of an interaction model between MCV and RBC, produces a marginally significant regression coefficient for hematocrit (P=.072). The pattern of positive and negative coefficients for MCV, RBC, and hematocrit is notably opposite of the pattern for the no-treatment group (1) and the nondiuretic-treated group (3), although interaction models in these groups do not produce statistically significant regression coefficients.
For DBP, the regression coefficient for MCV is also initially negative in all three groups and statistically significant (P=.015) for the no-treatment group (Table 2⇑). The addition of RBC as a covariate results in positive MCV regression coefficients, statistically significant in men taking any diuretic (group 2) and men taking nondiuretics only (group 3). Interaction models are statistically significant for the no-treatment group (1) and the nondiuretic-treated group (3) as reflected in the significance of the hematocrit coefficient. Both the sign and the magnitudes of the MCV, RBC, and hematocrit coefficients are comparable although slightly larger in the latter group.
The model predicting DBP and containing hematocrit as variables reflecting the MCV×RBC interaction is not significant (P=.28) for the group treated with any diuretic (group 2), although the model including only MCV and RBC does produce statistically significant coefficients for both of these variables. As with SBP, DBP systematically increases with increasing RBC, and a hint of flattening of the direct relation of blood pressure with MCV at higher RBC counts is noted (Fig 2⇓), although the statistics for an interaction model suggest that this is nothing more than random variation. Although not significant at α<.05, the patterns of positive and negative coefficients for the interaction model of MCV, RBC, and hematocrit (∝MCV×RBC) are completely opposite those noted for the no-treatment and the nondiuretic-treated groups (groups 1 and 3, respectively).
In aggregate, these findings suggest that relations among blood pressure and erythrocyte measures are different in men treated with diuretics, and it is not advisable to pool data from men in this group with data in men treated only with nondiuretics.
Pooled Group Relations
Men in groups 1 and 3 (no treatment and nondiuretics only) demonstrate similar patterns of regression relation. Accordingly, men in these groups are combined into a single group for analyses, although an indicator variable reflecting original group membership is added to models to take into account group differences in mean blood pressure and erythrocyte indexes (see Table 1⇑).
In this combined sampling frame, the models that include MCV alone indicate an inverse relation of SBP and DBP with MCV (Table 3⇓). The pattern of that relation among quartile groupings of MCV reveals markedly lower blood pressure at the highest MCV grouping for both measurements (Fig 3⇓). A statistical test that adds a squared term in MCV to the model suggests a significant relation for DBP (P=.05) but not SBP (P=.56). However, the addition of RBC as a covariate markedly attenuates the relation between SBP and MCV, and the model predicting DBP from MCV, RBC, and hematocrit produces statistically significant regression coefficients for all three erythrocyte measurements, with the MCV×RBC interaction variable hematocrit being the most prominent (P<.001).
The nature of relations between SBP and quartile groupings of RBC for the associations described in Table 3⇑ indicates a graded increase in blood pressure with increasing RBC, ≈5 mm Hg between the lowest and highest quartiles (Fig 4⇓). Within RBC quartile groupings, no relation with MCV is noted.
The nature of the same relation for DBP reveals a similar graded increase in blood pressure with increasing quartile group of RBC but also reveals the nature of the association of DBP with MCV and the interaction of MCV with RBC (Fig 5⇓). At the lowest level of RBC, there appears to be no relation between DBP and MCV. However, with increasing RBC, a direct relation becomes apparent such that at the highest levels of RBC, the DBP/MCV relation is at its steepest. Within the highest quartile grouping of RBC (≥5.0×1012/L), the difference in DBP between the lowest (<92.5 fL) and the highest (≥97.5 fL) MCV quartile groups is on the order of 3 mm Hg. This finding significantly contrasts with the suggestions of a flat blood pressure/MCV relation at higher RBCs among men treated with any diuretic (group 2) and a direct relation at lower RBCs (Figs 1⇑ and 2⇑). Too few observations and the attendant marginal statistical significance of the interaction models in this group (n=477) preclude any definitive assessment of this contrast.
Consistency of Findings
Without taking into consideration joint relations with other red cell measures, the MCV appears to be inversely related to SBP and DBP. This is demonstrated by the average SBP and DBP being higher and the average MCV being lower in group 3 (men treated exclusively with nondiuretic medication) than in group 1 (men with no treatment for hypertension) (Table 1⇑). Within these groups, both separately and after pooling, an inverse regression relation is also noted between SBP or DBP and MCV (Table 3⇑ and Fig 3⇑). It appears that this negative regression relation is primarily driven by data from the no-treatment group (Table 2⇑, MCV Alone). These results are consistent with the clinical studies reported by Bruschi et al1 and Postnov et al2 but inconsistent with the population-based studies of Strazzullo et al3 and Cirillo and Laurenzi.4
After adjustment for RBC, the MCV remains −0.49±0.16 fL lower (P=.002, not reported in tables) in men treated exclusively with nondiuretics (group 3) compared with men not treated (group 1) for hypertension (−0.85±0.18 fL without RBC adjustment, Table 1⇑). However, within each of these groups, the regression relation between blood pressure and MCV after adjustment for RBC is either nonexistent (eg, SBP) or positive (eg, DBP) (Tables 2⇑ and 3⇑). Thus, elements of both the clinical1 2 and the population-based3 4 studies are supported, depending on how the comparisons are structured.
Hypertension Treatment Effects
Observations among men treated with diuretics may be related to an effect of diuretics to reduce total plasma volume. Men treated with any diuretic had the highest hematocrit compared with the other two groups (Table 1⇑). Such “hemoconcentration” effects are well established.21 Simple correlations between MCV and RBC within the three groups (no treatment, any diuretic, and only nondiuretics) are remarkably comparable (−0.448, −0.453, and −0.448, respectively), although the SD of RBC (Table 1⇑) is significantly larger in men treated with any diuretic compared with the other two groups (0.49, 0.56 [P<.001], and 0.49×1012/L, respectively). Unfortunately, without measurements of total red cell mass and total plasma volume, contrasts of RBC and MCV among these treatment groups are difficult to interpret.
The aged nature of the Honolulu Heart Program population in this analysis may be a significant factor in explaining differences among the three treatment groups. Hemodynamic patterns of hypertensive disease in younger men are distinctly different than patterns noted in older men. Cardiac output is notably lower and total peripheral resistance is higher in older hypertensive men than in age-matched normotensives.23
Consistent among all three treatment groups is the relatively “strong” relation among RBC, SBP, and DBP. Population-based studies have consistently demonstrated direct relations between blood pressure and simple measures of erythrocyte properties, particularly hematocrit.8 9 10 11 12 13 14 15 16 The relation of hematocrit with blood pressure appears to be mediated by an impact on whole blood viscosity.16 24 25 Animal experiments indicate a causal relation between viscosity and both pulmonary and peripheral resistance to blood flow.26 Clinical studies in humans confirm this finding.27 28 Such a finding, although not necessarily sufficient to explain the relation between erythrocyte measures and blood pressure in studies such as this one, does suggest that a portion of that relation may be driven by fluid mechanical considerations whereby blood flow (U) to an organ system through a vessel of fixed diameter (d) and length (l) is maintained by an increase in blood pressure (P) when blood viscosity (η) becomes elevated, or (see Reference 2929 ):
A ratio of P/U reflects resistance to flow, thus implying that blood flow resistance is determined by blood viscosity and by “vascular hindrance,” this latter term representing the contribution of vascular geometry (d and l).30
Empirical equations for the viscosity of a suspension of particles (ηs) relative to the viscosity of the suspending phase (ηo) are defined in terms of the volume fraction of the particles in the total volume of the suspension (φ) by (see Reference 3131 ).
where ν and κ are empirically estimated coefficients, and higher orders of the volume fraction, φ, may be necessary to fully describe the relation. The volume fraction can be considered analogous to the hematocrit and, thus, blood viscosity changes nonlinearly with hematocrit.32 However, the relation between whole blood viscosity and the constituents of blood is complex and can be affected by red cell deformability, reversible aggregation between red cells, and flow characteristics in the vasculature.30 In the present study, we chose to conceptualize red cell “mass” as RBC in the belief that cell “count” and “volume” variables more specifically measure underlying physiological processes. Although the contribution of the red cell mass to whole blood viscosity can be thought of in terms of the hematocrit and although the hematocrit is proportional to the product of MCV×RBC, the rheological behavior of blood is sufficiently complicated to preclude the expectation that a constant hematocrit implies a constant viscosity. In fact, recent experimental work demonstrates significant decreases in whole blood viscosity with decreasing MCV at a constant hematocrit,33 thus indicating the need to consider both MCV and RBC rather than only their product.
Within the foregoing conceptual framework, the following model is suggested by these results.
(1) Because essential hypertension is primarily a derangement of peripheral vascular resistance30 and because DBP is a more specific measure of overall resistance to blood flow than SBP, blood characteristics that influence viscosity will be more strongly related to DBP than to SBP.
(2) In general populations, there is a sufficiently high prevalence of and/or variation in health conditions that increase relative red cell mass to the point of affecting hemodynamic function.
(3) The impact of an increase in relative red cell mass is to increase whole blood viscosity, primarily by increasing the number of particles per unit volume of blood (RBC), and thereby increase peripheral resistance to blood flow.
(4) To maintain blood flow in the face of increased peripheral resistance, blood pressure increases.
(5) The deleterious consequence of increasing pressure, presumably to maintain blood flow, is partially compensated for by a concomitant decrease in red cell volume, thus attempting to counteract the viscous effects of a larger relative red cell mass (ie, count) with smaller cell size characteristics.
Because blood pressure level would appear to be a necessary modulator in this hypothesis, an interaction is implied of pressure regulation with counterregulatory relations between RBC and MCV. This is the key implication of the results presented in Fig 3⇑ and is partially supported by laboratory-based research via rheological mechanisms.33 The hypothesis attempts to integrate known physiological processes linking blood pressure, blood flow, and peripheral resistance to flow with known relations involving fluid mechanics and viscosity of whole blood. It does not purport to describe how viscosity characteristics, via red cell mass, come to be altered, nor whether such alterations precede or follow a derangement in blood pressure regulation.
Clinical implications ensue from this model. Although early hypertensive disease is associated with increased resistance to flow, this appears to be primarily mediated by increases in blood viscosity rather than vascular hindrance.30 However, in older men or men with more severe hypertensive disease, both components contribute to increased resistance, and, indeed, both could be linked as a manifestation of perturbed pressure/flow regulation. Because of the geriatric nature of this study group, variation in viscosity characteristics may play a relatively more important role than vascular hindrance in the observed relations with DBP.16 Vessel geometry, vessel compliance, and end-arteriole reactivity may be relatively static due to atherosclerotic processes and the consequences of chronic hypertension in this elderly population compared with hypertensive men 30 to 40 years younger. Thus, if the model implied by this hypothesis is true, then maintenance of blood flow to critical organ systems may be better served by factors that decrease blood viscosity with little compromise of oxygen-carrying capacity than by factors that alter vessel hindrance.
The model also conveys public health implications. The results suggest that when there is a sufficiently high prevalence of a pathophysiological process, some kind of regulatory interrelationship may be created, such as in this case among DBP, RBC, and MCV. One such example of a pathophysiological condition in an elderly population may be chronic obstructive lung disease and its possible sequel of polycythemia.7 Whether these interrelations would be manifest in other populations with different age, sex, and ethnic characteristics may depend on how such demographic characteristics are related to the existence and manifestation of similar disease processes affecting blood pressure control and the regulation of relative red cell mass. Longitudinal studies in particular may contribute unique understanding as to how derangements in these physiological systems coevolve and are related to the incidence of overt clinical disease.
Selected Abbreviations and Acronyms
|DBP||=||diastolic blood pressure|
|MCV||=||mean corpuscular volume|
|RBC||=||red blood cell count|
|SBP||=||systolic blood pressure|
This work was supported by contract N01–HC–05102 (to the Honolulu Heart Program, Kuakini Medical Center, Honolulu, Hawaii) from the National Heart, Lung, and Blood Institute; National Institutes of Health Research Grants HL–15722 and HL–48484 (Drs Meiselman and Fisher); and an award from the Wright Foundation, Los Angeles, Calif (Dr Fisher).
- Received July 24, 1995.
- Revision received October 23, 1995.
- Accepted November 3, 1995.
- Copyright © 1996 by American Heart Association
Bruschi G, Minari M, Bruschi ME, Tacinelli L, Milani B, Cavatorta A, Borghetti A. Similarities of essential and spontaneous hypertension: volume and number of blood cells. Hypertension. 1986;8:983-989.
Postnov YV, Kravtsov GM, Orlov SN, Pokudin NI, Postnov IY, Kotelevtsev YV. Effect of protein kinase C activation on cytoskeleton and cation transport in human erythrocytes: reproduction of some membrane abnormalities revealed in essential hypertension. Hypertension. 1988;12:267-273.
Cirillo M, Laurenzi M. Erythrocyte and platelet volume in human hypertension. J Hypertens. 1989;7(suppl 6):S168-S169.
Gaisböck F. Die Bedeutung der Blutdruck-messung für die ärztliche Praxis. Deutsch Arch Klin Med. 1905;83:363-409.
Harlan WR, Hull AL, Schmouder RL, Landis JR, Thompson FE, Larkin FA. Blood pressure and nutrition in adults: the National Health and Nutrition Examination Survey. Am J Epidemiol. 1984;120:17-28.
Havlik RJ, Garrison RJ, Feinleib M, Padgett S, Castelli WP, McNamara PM. Evidence for additional blood pressure correlates in adults 20–56 years old. Circulation. 1980;61:710-715.
Fowkes FHR, Lowe GDO, Rumley A, Lennie SE, Smith FB, Donnan PT. The relationship between blood viscosity and blood pressure in a random sample of the population aged 55 to 74 years. Eur Heart J. 1993;14:597-601.
Yano K, Reed DM, Mcgee DL. Ten–year incidence of coronary heart disease in the Honolulu Heart Program: relationship to biologic and lifestyle characteristics. Am J Epidemiol. 1984;119:653-666.
Cirillo M, Laurenzi M, Trevisan M, Stamler J. Hematocrit, blood pressure, and hypertension: the Gubbio population study. Hypertension. 1992;20:319-326.
Leth A. Changes in plasma and extracellular fluid volumes in patients with essential hypertension during long–term treatment with hydrochlorothiazide. Circulation. 1970;42:479-485.
SAS Institute Inc. SAS Companion for the Microsoft Windows Environment, version 6. 1st ed. Cary, NC: SAS Institute Inc; 1993:356.
Lund–Johansen P, Omvik P. Hemodynamic patterns of untreated hypertensive disease. In: Laragh JH, Brenner BM, eds. Hypertension: Pathophysiology, Diagnosis, and Management. New York, NY: Raven Press Ltd; 1990:305-327.
de Simone G, Devereux RB, Chien S, Alderman MH, Atlas SA, Laragh JH. Relation of blood viscosity to demographic and physiologic variables and to cardiovascular risk factors in apparently normal adults. Circulation. 1990;81:107-117.
Letcher RL, Chien S, Pickering TG, Laragh JH. Elevated blood viscosity in patients with borderline essential hypertension. Hypertension. 1983;5:757-762.
Berne RM, Levy MN. Cardiovascular Physiology. New York, NY: Mosby-Year Book; 1992:120.
Chabanel A, Chien S. Blood viscosity as a factor in human hypertension. In: Laragh JH, Brenner BM, eds. Hypertension: Pathophysiology, Diagnosis, and Management. New York, NY: Raven Press Ltd; 1990:329-337.
Van Holde KE. Physical Biochemistry. Englewood Cliffs, NJ: Prentice–Hall Inc; 1971:145.
Lowe GDO. Clinical Blood Rheology. Boca Raton, Fla: CRC Press; 1988:27.