214 Volume 6 - Number 2 #{149}1995

Determination of Circulating Blood Volume byContinuously Monitoring Hematocrit During ,2

John K. Leypoldt,3 Alfred K. Cheung, Robert R. Steuer, David H. Harris, and James M. Conis

J.K. Leypoldt. Research Service. VA Medical Center.Salt Lake City. UT

AK. Cheung. Medical Service. VA Medical Center.Salt Lake City. UT

J.K. Leypoldt. Departments of Medicine and Bioengi-neering. University of Utah, Salt Lake City, UT

AK. Cheung. Department of Medicine. University ofUtah. Salt Lake City, UT

R.R. Steuer. D.H. Harris. J.M. Conis. In-Line DiagnosticsCorporation. Riverdale. UT

(J. Am. Soc. Nephrol. 1995; 6:214-2�9)

ABSTRACTDialysis-induced hypovolemia occurs because the

rate of extracorporeal ultrafiltration exceeds the rateof refilling of the blood compartment. The purpose ofthis study was to evaluate a method for calculatingcirculating blood volume (BV) during hemodialysis

(HD) from changes in hematocrit(Hct) shortly (2 to 10mm) before and after ultrafiltration (UF) was abruptly

stopped. Hct was monitored continuously during 93HD treatment sessions in 16 patients by an opticaltechnique and at selected times by centrifugation of

blood samples. Total plasma protein and albuminconcentrations were also measured at selectedtimes. Continuously monitored Hct correlated with Hct

determined by centrifugation (R - 0.89, N = 579).Relative changes in BV determined by continuouslymonitored Hct were not different from those deter-

mined by total plasma protein concentration (P =

0.05; N = 273). Calculated BV at the start of dialysis(4.1 ± 1.3 L) was not different (P = 0.18, N = 12) fromthat derived anthropometrically from the patient’s dry

weight (4.6 ± 0.8 L), and calculated BV when UF wasstopped was 3.2 ± 0.5 L (46 ± 7 mI/kg body wt). Theselafter estimates of BV are consistent with those deter-mined previously by dilution techniques in HD pa-tients. It was concluded that( 1) relative changes in BVassessed by continuously monitored Hct were unbi-ased and (2) BV can be determined noninvasively

1 Received May 18, 1994. Accepted May 17. 1995.2 � 5teuer, D.H. Harris, J.M. conis are employees of In-line Diagnostics Corpo-

ration.

3 Correspondence to Dr. J.K. Leypoldt, VA Medical Center (1 1H). 500 FoothIllBlvd.. salt Lake City. UT 84148.

1046’6673/0602-0214$03.00/0Journal of the American Society of NephrologyCopyright C 1995 by the American society of Nephrology

during HD by continuously monitoring Hct and tem-

porarily stopping UF.

Key Words: Blood volume, hematocrit, hemodialysis, monitor

D espite the ability of commercial hemodialysis

machines to accurately remove fluid at a given

desired rate, intradialytic hypovolemia and hypoten-sion routinely occur because of an imbalance between

the rate ofextracorporeal ultrafiltration and the rate of

refilling of the blood compartment ( 1 ). Previous work

has shown that the refilling rate is patient specific and

depends on both Starling forces within the blood

compartment (2,3) and the hydration status of inter-

stitial tissue (4). To assist in the prevention of intra-

dialytic hypotension, several devices have been re-

cently developed to monitor changes in blood volume

during hemodialysis (5-9). These devices monitor ei-

then hematocnit (Hct) (usually by assessing blood he-

moglobin concentration) or plasma protein concentra-

tion, therefore permitting the calculation of relative

changes in blood volume (BV) on the basis of the

principle of mass conservation. This approach is at-

tractive because it can be adapted for continuousnoninvasive monitoring; however, it can only deter-

mine relative changes in, but not absolute values of,circulating BV.

Schallenberg et at. ( 10) proposed that absolute cm-

culating BV could be determined from relative

changes in BV (assessed by blood hemoglobin concen-tration) before and after ultrafiltration (UF) was tem-

poraily stopped. They derived an equation for calcu-

lating circulating BV but reported results for only one

patient. The calculated values of circulating BV were

variable and less than expected on the basis of the

patient’s body weight; thus, these investigators con-

cluded that hemoglobin is likely distributed unevenly

in the intravascular space. The purpose of this study

was to evaluate the method of Schallenberg et at. (10)

for calculating circulating BV during multiple clinical

hemodialysis sessions from changes in Hct shortly

before and after abruptly stopping UF. We found that

calculated values of circulating BV were reasonable

and consistent with those previously reported by the

use of other methods.

METHODS

Experimental

Sixteen (5 male and 1 1 female) patients at the University ofUtah-affiliated Bonneville Dialysis Unit (Ogden, UT) wereenrolled in this study after giving informed consent. Each

patient was studied on six separate occasions for a total of 93

treatment sessions. (One patient missed two sessions be-

cause he received a renal transplant. and one patient missed

one session because she had a clotted vascular access.)

Patients were treated by routine dialysis with minimal equation can be derived (see Appendix) for calculating circu-

intervention by the research staff. Hemodialysis was per- lating BV at the start of dialysis, BV(0)formed with 2008E machines (Fresenius, Concord, CA), bi-carbonate dialysate, and cellulose acetate membrane hollow-fiber dialyzers (CA series; Baxter Healthcare. Round Lake,

E ti 2( qua on

IL). The dialysate flow rate was fixed at 500 mL/min, but theblood flow rate and treatment time were prescribed individ-ually for each patient by the use of urea kinetics. The UF rate where T- and T+ denote the times shortly before and after,

(UFR) ranged between 457 and 1 .971 mL/h and was selected respectively, UF is stopped. Circulating BV at the time UF isby the dialysis staff to remove sufficient fluid to achieve the stopped, BV(r), can also be calculated by an analogouspatient’s prescribed dry weight. The predetermined UFR was equation. We found it most convenient to calculate this latterheld constant during each session, except when It was parameter directly from changes In Hct (see Appendix) as

necessary to change the rate because of intradialytic symp-toms. The dialysate sodium concentration was programmed

to vary linearly from 150 mEq/L at the start to 142 mEq/L at�E ti 3‘ qua on

the end of dialysis for all sessions.Hct was monitored noninvasively and continuously with

the CRIT-LINE instrument (In-Line Diagnostics, Riverdale, Equations 2 and 3 demonstrate that it is possible to calculateUT) during each session, as described (8). Before hemodial- circulating BV during hemodlalysis when UF is temporarilyysis, a sterile, plastic, disposable blood chamber (Beta pro- stopped by determining the rate of change (d/dt) of Hct (ortotype; In-Line Diagnostics) was placed in the blood circuit BV) shortly before and after stopping UF. It should bebetween the arterial blood tubing and the dialyzer. The emphasized that these relationships are not valid when otherCRIT-LINE instrument uses a transmissive photometric interventions (e.g. , saline infusion, use of the Trendelenburgtechnique to determine the Hct on the basis of both the position), in addition to stopping UF, are also performedabsorption properties ofhemoglobin and the scattering prop- because they will alter Hct and BV (10, 1 1).erties of red blood cells passing through the blood chamber. Both hematocnit-time and �BV-tlme profiles were plottedThe tubing set, disposable blood chamber, and blood com- for each treatment session. From the �BV-time profile, cm-partment of the dialyzer were then rinsed and primed with culating BV at the start ofhemodialysis was calculated by thenormal saline. Although dialyzers were reused, the blood use of Equation 2, and from the Hct-time proffle, circulatingtubing and blood chamber were not. Blood samples were BV at the time UF was stopped was calculated by the use oftaken before hemodialysis, hourly during the session. and Equation 3. The derivatives in Equations 2 and 3 wereimmediately before stopping hemodialysis for determining determined by drawing a straight line visually through the

Hct by microcentrifugatlon and total plasma protein and proffles 2 to 10 mm before and after UF was stopped (seealbumin concentrations. During one session on each patient, Figure 3B below).blood samples were also taken every 1 5 mm for determiningHct by microcentrifugatlon. Analytical

Calculations

Total plasma protein and albumin concentrations were

determined with an automated analyzer (SYNCHRON CX-5 &CX-7; Beckman, Brea, CA). Intra-assay precision for total

The relative change in blood volume (�BV) was calculatedfrom the observed change in Hct by use of the following

plasma protein determinations of 0.3 g/dL (standard devia-tion) was reported by the manufacturer.

equationStatistics

(E uation 1 )q

Linear regression was performed with Lotus 1-2-3 Release

� software (Cambridge, MA). All values are reported as the

mean value ± the standard deviation. Analyses with P < 0.01The values of Hct and BV at the start of hemodialysis are were considered significant.denoted as H(0) and BV(0), respectively; in this and subse-quent equations, Hct is denoted as H. Although H and �BV RESULTScan be determined at any time during hemodialysis, absolutevalues of circulating By, I.e., BV and BV(0), cannot. H and�BV were recorded and calculated, respectively, by the CRIT-LINE instrument every 10 s during hemodialysis. Hct values

Although determinations of Hct by the CRIT-LINE

instrument correlated well with those determined by

centnl.fugatlon (Figure 1 ), the slope ( 1 .3 1 ± 0.03) wasare expressed as a percentage of total BV; the units are not equal to 1 (p < 0.00 1 ). By the use of a nestedimplied and not specifically indicated. analysis of variance statistical model ( 1 2), it was

Assuming that changes in BV during hemodialysis occuronly because of extracorporeal UF and vascular refilling, BV

will decrease when the UFR exceeds the vascular refillingrate. Conversely, BV will increase as the result of vascularrefilling when UF is stopped temporarily during hemodialy-sis. Suppose that UF is abruptly stopped at a certain time,denoted by ‘t-, during hemodlalysis. Shortly before T, BV will

be decreasing, and shortly after ‘r, BV will be increasing. If itIs assumed that the vascular reffiling rate Is identical shortly

shown that discrepancies between Hct estimates were

larger between different treatment sessions than

within sessions (P < 0.00 1 ). The standard deviation ofthe difference in Hct estimates within treatment ses-

5i0n5 was 1 .6, whereas that between sessions was 2.4.

This analysis suggests that the CRIT-LINE instrumentwas able to detect relative changes in Hct within an

individual treatment session better than that implied

before and after r, before the body has time to adjust, an from the results shown in Figure 1.

Leypoldt et al

Journal of the American Society of Nephrology 215

BV H(0)z��BV = BV(0) � � = �

UFR

BV(0) = d(z�BV) d(�BV)

dt � dt (T-)

H(T)UFR

BV(T)=�H dH

�-(T-) -

wz-J

0

8CuEa)I

A5

0

-5

< -10

-15

-200

30 40

Hematocrit by Centrifugation

I 2 3 4

�flme (his)

B1#{176}

0

-10

-20

-300 I 2 3 4

wz-J

0>,

.0

>

-30 -20 -10 0

�BV by Plasma Protein (%)

Blood Volume During Hemodialysis

216 Volume 6 . Number 2 . 1995

Figure 1 . Hct determined by the CRIT-LINE instrument plottedversus Hct determined by centrifugation. All datum pointsfrom 93 treatment sessions are shown. The line shown is theidentity line.

Relative changes in BV determined by the CRIT-

LINE instrument also correlated well with those deter-

mined by total plasma protein concentration (Figure

2). Although there is considerable scatter in this plot,

there was no difference (0.9 ± 6.0%; P = 0.05) between

these two estimates of �BV when they were compared

with a paired t test. This analysis implies that �BV

determined by the CRIT-LINE instrument and by total

plasma protein concentration give indistinguishableresults. Relative changes in BV determined by the

CRIT-LINE instrument also correlated well with those

determined by albumin concentration (R = 0.64, N =

261; graph not shown).Figures 3A and B show examples of relative changes

in BV determined by the CRIT-LINE instrument dun-

ing treatment sessions on different patients. In the

Figure 2. Relative changes in BV (�BV) determined by theCRIT-LINE instrument plotted versus that determined by totalplasma protein concentrations. All datum points from 93treatment sessions are shown. The diagonal line shown is theidentity line.

Time (hrs)

Figure 3. (A) Relative changes in BV (�XBV) during hemodial-ysis for one sample treatment session. The UFR of 800 mL/hwas held constant throughout this session. (B) Relativechanges in BV (�BV) during hemodialysis for another sam-pIe treatment session. UF was conducted at 1 200 mL/h butwas abruptly stopped (UF off) because of muscle cramps.The method for determining thBVIdt(T-) and thBV/dt(T+)is also illustrated here.

example shown in Figure 3A, the decrease in BV wasmost rapid at the beginning of treatment. The shape of

this �BV-time profile is similar to that predicted by

mathematical models previously described (2,3). Pro-files of this shape were not common in this study,

however. A more typical profile is shown In Figure 3B,where relative changes in BV were minimal at the

beginning of treatment but where BV decreased sub-

stantially later in the session. In this example, UF was

stopped temporarily when this patient experienced

muscle cramps. The method for determining d(z�BV)/

dt(T-) and d(z�BV)/dt(T+) is also shown on this figure.

Circulating BV was determined when UF was

stopped during 32 treatment sessions on 12 patients

who suffered Intradialytic morbidity such as hypoten-

sion, muscle cramps, or lightheadedness and no ad-

ditional interventions besides stopping UF were per-

formed. Figure 4 shows the average values of

-J

0

>

a)

E

0>�000

a)C

Cu

0

0

Anthropometric Blood Volume (L)

0)

E

>

a)

E

00

m0)

C

Cu

0

0

60 #{149} .

50

. S #{149}

40 .

I

30 #{149}

20

10

Leypoldt et al

Journal of the American Society of Nephrology 217

Figure 4. Circulating BV at the start of hemodialysis, BV(O),calculated by the use of Equation 3 plotted versus anthro-pometric BV derived from the patient’s dry body weight (13).Only mean values for each patient are shown.

circulating BV at the stat of hemodialysis for each

patient plotted versus anthropometric BV derived

from the patient’s dry weight as described for nonure-

mic volunteers by Boer ( 1 3). The values calculated bythe use ofEquation 2 (4. 1 ± 1 .3 L) were not different (P= 0. 18) from those derived from the patient’s dry

weight (4.6 ± 0.8 L). Figure 5 shows each individual

calculated value of circulating BV at the time UF was

stopped for each patient. The average value for all 12patients was 3.2 ± 0.5 L (46 ± 7 mL/kg body wt).

DISCUSSION

Previous work has demonstrated that relativechanges in BV can be determined during routinehemodialysis therapy by continuously monitoring Hct

Th 2 4 6 8 10 12 14

Patient Number

Figure 5. Circulating BV at the time when extracorporeal UFwas stopped because of intradialytic morbidity, BV(T), cal-culated by the use of Equation 4 for all determinations (N =

32) on the 12 patients. Circulating BV was normalized by thepatient’s dry body weight and was estimated between oneand five times on the patients.

or plasma protein concentrations (5-9). When using

Hct to assess relative changes in By, it is assumed

that there are no changes in red blood cell volumeduring hemodialysis. Previous work has suggestedthat changes in red blood cell volume during hemodi-alysis are indeed small despite changes in plasmasodium concentration and plasma osmolality (14-16).In this study, Hct determined by the CRIT-LINE in-

strument correlated well with those determined bycentnifugation but not as well as with those deter-

mined by this instrument in vitro (R = 0.996) (8). Our

statistical analysis demonstrated that some of thedifferences between the Hct determined by the CRIT-LINE instrument and that determined by centnifuga-

tion occurred between treatment sessions and couldbe ascribed to additional variability associated with

the disposable blood chambers or, possibly, with cen-tnifugal determinations of Hct.

The good correlation between relative changes in BVassessed by the CRIT-LINE instrument and those

assessed by total plasma protein concentration (andalbumin concentration) demonstrates that these ap-

proaches produce similar results. Although there isconsiderable scatter between these estimates of rela-

tive change in BV (Figure 2), it should be noted thatthe standard deviation between these estimates

(6.0%) is comparable to that for determinations of

total plasma protein concentration by the automated

analyzer. Therefore, we can only conclude that relativechanges in BV assessed by the CRIT-LINE instrument

are unbiased and at least as accurate as those as-sessed by changes in total plasma protein concentra-tion measured by the use of routine clinical assays.

The temporal profile of relative changes in BV deter-

mined in this study were highly variable, and thisvariability likely reflects patient-specific characteris-tics in response to the acute removal of large amountsof fluid from the blood compartment during hemodi-alysis. Previous mathematical models of body fluidkinetics during hemodialysis have suggested that BVshould decrease most rapidly at the beginning of

therapy (2,3). The results from this study did not

confirm those predictions. This lack of agreement

between theory and experimental results suggeststhat factors other than those included in previousmathematical models, such as the variable dialysatesodium concentration used, the hydration status ofInterstitial tissue (4) and physiologic responses affect-

ing the heat and peripheral circulation ( 1 7, 1 8), are

likely important in governing changes in BV duringhemodialysis.

Schneditz et at. (3) have recently calculated values

of absolute BV in hemodialysis patients by comparingrelative changes in BV In response to a short pulse of

UF during the first hour of hemodialysis with that

predicted by a mathematical model. These calculated

values of BV correlated well with those calculated fromanthropometry. We used a different approach in thisstudy to estimate circulating BV because the shape ofmost �BV-time profiles did not correspond to those

Blood Volume During Hemodialysis

218 Volume 6 ‘ Number 2 . 1995

predicted from previous mathematical models. This

approach for estimating circulating BV is similar tothat previously attempted on a single patient by

Schallenbeng et at. ( 10) and is model independent,relying instead on the assumption that the vascular

reffiling rate is not significantly altered over severalminutes when UF is stopped. Although this assump-tion has not yet been experimentally validated, it

appears reasonable and there Is no experimental evi-

dence to the contrary. These results on 12 patients aremore favorable than those reported previously and

suggest that this approach may provide a method for

noninvasively determining circulating BV in hemodi-

alysis patients. We were unable to demonstrate a

significant difference between circulating BV at thestat of hemodialysis and those calculated from the

patient’s dry weight by anthropometry. This result

was not different if either the patient’s predialysis

body weight or the patient’s lean body mass (as cal-

culated from both height and weight I 131) were insteadused as the basis for calculating circulating BV (data

not shown).

The values of circulating BV reported in this study

can also be compared with values of BV previously

determined in hemodialysis patients by the dilution

techniques of Kim et at. ( 1 9). These investigatorsdetermined BV before the stat of hemodialysis in

three groups of patients and calculated mean BV of3.817, 4.429, and 4.058 L. Moreover, they reported

that BV less than 2.8 L (50 mL/kg body wt) were

associated with Intradialytic hypotension. Our esti-

mates of circulating BV are consistent with those

previous results.Although detailed physiologic responses to volume

reduction were not considered in this study, we have

recently reported that relative changes in BV corre-

lated with the amount of fluid removed during hemo-

dialysis and that Intradialytic morbid events occurredwhen the Hct reached a patient-specific threshold

(20). The latter observation implies that intradialytic

morbid events occurred when the absolute circulatingBV decreased to a critical value. The capability of

determining absolute circulating BV by the technique

described in this study would allow this concept to be

applied even when the red blood cell mass changesover time. This and other potential clinical applica-

tions of absolute circulating BV measurements, how-

ever, need further validation.

We conclude that relative changes in BV assessed

by continuously monitored Hct were unbiased and

that circulating BV during hemodialysis calculated by

changes in continuously monitored Hct before andafter temporarily stopping UF were not different from

anthropometric estimates and were similar to those

previously determined by dilution techniques.

ACKNOWLEDGMENTS

The authors thank Dr. Harry 0. SenekJian. David Dalpias. and the

staff and patients of the Bonneville Dialysis Center for their partici-

pation in this study. This work was supported by In-Line Diagnostics

Corporation and DVA Medical Research Funds.

APPENDIX

Although Equation 2 was derived previously bySchallenberg et at. ( 10), an alternative derivation is

shown here to clarify present terminology. Changes incirculating BV with time [d(BV)/dtl during hemodial-ysis are dependent on the UFR and the vascular

refilling rate (VRR) according to the following equation

d(BV)

- dt UF’R - VRR (Equation 1A)

Suppose that at time T, UF is stopped abruptly during

hemodialysis. Shortly before UF is stopped, the circu-lating BV will decrease according to Equation 1A, or

d(BV)

- dt (T ) = UFR - VRR (Equation 2A)

where T denotes the time shortly before UF is

stopped. Shortly after stopping UF, circulating BV willIncrease according to Equation 1A with UFR set equal

to zero, or

d(BV)

dt (T + ) = VRR (Equation 3A)

where T+ denotes the time shortly after UF is stopped.

The relationship between changes in circulating BV

and changes in L�BV is (see Equation 1 in the Text)

d(BV) d(LIBV)

dt BV(0) dt (Equation 4A)

where BV(0) is the value of circulating BV at thebeginning of hemodialysis. Assuming that the VRR is

identical shortly before and after UF is stopped, VRR

can be eliminated from Equations 2A and 3A and the

result for BV(O) can be solved using Equation 4A as

UFRBV(O) = d(�BV) d(�BV)

dt � � dt (iS-)

(Equation 5A)

This is Equation 2 In the Text.The relationship between changes in �BV and

changes in Hct (H) is (see Equation 1 in the Text)

d(z�BV) H(O) dli

dt = � � � (Equation 6A)

Substituting Equation 6A into Equation 5A and using

Equation 1 in the Text yields an equation for circulat-ing BV at the time UF is stopped

H(T)UFRBV(r)=� dH

This is Equation 3 in the Text.

(Equation 7A)

Leypoldt et al

Journal of the American Society of Nephrology 219

REFERENCES

1 . Levin NW, Kupin WL, Zasuwa G, Venkat KK: Complica-tions during hemodialysis. In: Nissenson AR, Fine RN,Gentile DE, Eds. Clinical Dialysis. 2nd Ed. Norwalk:Appleton & Lange; 1990:172-201.

2. Keshaviah PR, listrup KM, Shapiro FL: Dynamics ofvascular refilling. In: Atsumi K, Maekawa M, Ota K, Eds.Progress in Artificial Organs-1983. Cleveland: ISAOPress: 1984:506-510.

3. Schneditz D, Robb J, Oswald M, et aL: Nature and rateof vascular refilling during hemodialysis and ultrafiltra-tion. Kidney Int 1992;42: 1425-1433.

4. Koomans HA, Geers AB, Mees EJD: Plasma volumerecovery after ultrafiltration in patients with chronicrenal failure. Kidney Int 1984;26:848-854.

5. Stiller S, Th#{246}mmes A, Konigs F, Schallenberg U, MannH: Characteristic profiles of circulating blood volumeduring dialysis therapy. Trans Am Soc Artif Intern Or-gans 1989;35:530-532.

6. Schneditz D, Pogglitsch H, Horina J, Binswanger U: Ablood protein monitor for the continuous measurementofblood volume changes during hemodialysis. Kidney Int1990;38:342-346.

7. de Vries J-PPM, Olthof CG, Visser V, et aL: Continuousmeasurement ofblood volume during hemodialysis by anoptical method. ASAIO J 1992;38:M181-M185.

8. Steuer RR, Harris DH, Conis JM: A new optical tech-nique for monitoring hematocrit and circulating bloodvolume: Its application in renal dialysis. Dial Transplant1 993;22:260-265.

9. Mancini E, Santoro A, Spongano M, Paolini F, Rossi M,Zucchelli P: Continuous on-line optical absorbance re-cording of blood volume changes during hemodialysis.Artif Organs l993;17:691-694.

10. Schallenberg U, Stiller S, Mann H: A new method of

continuous haemoglobinometric measurement of bloodvolume during hemodialysis. Life Support Sys 1987:5:293-305.

1 1 . Steuer RR, Harris DH, Conis JM: Instantaneouschanges in circulating blood volume due to variousphysiological maneuvers. Dial Transplant l994;23:643-646, 650.

12. Sokal RR, Rohif FJ: Biometry. 2nd Ed. San Francisco:WH Freeman; 1981:271-320.

13. Boer P: Estimated lean body mass as an index fornormalization of body fluid volumes in humans. Am JPhysiol 1984;247:F632-F636.

1 4. Swartz 1W, Somermeyer MG, Hsu C-H: Preservation ofplasma volume during hemodlalysis depends on dialy-sate osmolality. Am J Nephrol 1982:2:189-194.

15. Hsu CH, Swartz PD, Somermeyer MG, Raj A: Bicarbon-ate hemodialysis: Influence on plasma refilling and he-modynamic stability. Nephron 1984;38:202-208.

16. Fleming SI, Wilkinson JS, Aidridge C, et at. : Dialysis-induced change in erythrocyte volume: Effect on changein blood volume calculated from packed cell volume. ClinNephrol 1988;29:63-68.

1 7. Daugirdas JT: Dialysis hypotension: A hemodynamicanalysis. Kidney Int 199139:233-246.

18. Converse RL Jr, Jacobsen TN, Jost CMT, et at.: Para-doxical withdrawal of reflex vasoconstriction as a causeof hemodialysis-induced hypotension. J Clin Invest1992;90: 1657-1665.

19. Kim KE, NeffM, Cohen B, et at.: Blood volume changesand hypotension during hemodialysis. Trans Am SocArtif Intern Organs 1970; 16:508-514.

20. Steuer RR, Leypoldt JK, Cheung AK, Harris DH, ConisJM: Hematocrit as an indicator of blood volume and apredictor of mntradialytic morbid events. ASAIO J 1994;40:M69 1-M696.