Weiskopf RB, Viele MK, Feiner J, Kelley S, Lieberman J, Noorani M, Leung JM, Fisher DM, Murray WR, Toy P, Moore MA. Human Cardiovascular and Metabolic Response to Acute, Severe Isovolemic Anemia. JAMA. 1998;279(3):217-221. doi:10.1001/jama.279.3.217
From the Departments of Anesthesia (Drs Weiskopf, Feiner, Kelley, Lieberman, Leung, Fisher, and Moore and Ms Noorani) and Physiology (Dr Weiskopf), Laboratory Medicine (Drs Viele and Toy), Orthopaedic Surgery (Dr Murray), and the Cardiovascular Research Institute (Dr Weiskopf), University of California, San Francisco.
Context.— Although concern over the risks of red blood cell transfusion has resulted
in several practice guidelines for transfusion, lack of data regarding the
physiological effects of anemia in humans has caused uncertainty regarding
the blood hemoglobin (Hb) concentration requiring treatment.
Objective.— To test the hypothesis that acute isovolemic reduction of blood Hb concentration
to 50 g/L in healthy resting humans would produce inadequate cardiovascular
compensation and result in tissue hypoxia secondary to inadequate oxygen transport.
Design.— Before and after interventional study.
Setting.— Academic tertiary care medical center.
Participants.— Conscious healthy patients (n=11) prior to anesthesia and surgery and
volunteers not undergoing surgery (n=21).
Interventions.— Aliquots of blood (450-900 mL) were removed to reduce blood Hb concentration
from 131 (2) g/L to 50 (1) g/L [mean (SE)]. Isovolemia was maintained with
5% human albumin and/or autologous plasma. Cardiovascular parameters, arterial
and mixed venous oxygen content, oxyhemoglobin saturation, and arterial blood
lactate were measured before and after removal of each aliquot of blood. Electrocardiogram
and, in a subset, Holter monitor were monitored continuously.
Main Outcome Measures.— "Critical" oxygen delivery (TO2) as assessed by oxygen consumption
(O2), plasma lactate concentration, and ST changes on electrocardiogram.
Results.— Acute, isovolemic reduction of Hb concentration decreased systemic vascular
resistance and TO2 and increased heart rate, stroke volume, and
cardiac index (each P<.001). We did not find evidence
of inadequate oxygenation: O2 increased slightly from a mean
(SD) of 3.07 (0.44) mL of oxygen per kilogram per minute (mL O2·kg−1·min−1) to 3.42 (0.54) Ml O2·kg−1·min−1 (P<.001)
and plasma lactate concentration did not change (0.81 [0.11] mmol/L to 0.62
[0.19] mmol/L;P=.09). Two subjects developed significant
ST changes on Holter monitor: one apparently related to body position or activity,
the other to an increase in heart rate (at an Hb concentration of 46-53 g/L);
both occurred in young women and resolved without sequelae.
Conclusions.— Acute isovolemic reduction of blood Hb concentration to 50 g/L in conscious
healthy resting humans does not produce evidence of inadequate systemic TO2, as assessed by lack of change of O2 and plasma lactate
concentration. Analysis of Holter readings suggests that at this Hb concentration
in this resting healthy population, myocardial ischemia would occur infrequently.
THE MOST COMMON indication for the transfusion of red blood cells is
to augment the oxygen-carrying capacity of the blood of an anemic patient.
The potential for blood components to transmit infectious disease has produced
concern and "guidelines," "practice parameters," or "standards" delineating
conditions under which it is appropriate to transfuse red blood cells.1- 4 Formulation
of these documents has been hindered by the lack of data defining the hemoglobin
(Hb) concentration in humans that does not allow for adequate oxygen transport
(TO2) and that initiates tissue hypoxia.
Accordingly, we acutely decreased the Hb concentration of 32 healthy
resting humans to test the hypothesis that a blood Hb concentration of 50
g/L would result in tissue hypoxia secondary to inadequate TO2.
We did not conduct these studies with the subjects performing any activity
that would thereby increase oxygen consumption (O2). Consequently,
our results are applicable only to humans at rest.
With approval of our institutional review board and with informed consent,
we studied 11 patients before a surgical procedure with an anticipated blood
loss of 2 L or more and 21 paid volunteers not undergoing surgery. Data from
patients were collected before anesthesia and surgery. All volunteers and
patients were without cardiovascular, pulmonary, or hepatic disease; did not
smoke; and were not taking any drugs with cardiovascular actions.
Peripheral venous, radial arterial, and flow-directed pulmonary arterial
(via the right internal jugular vein; Baxter Healthcare, Glendale, Calif)
cannulas were inserted into each subject using local anesthesia. In 18 subjects
an intravenous infusion of propofol (50-150 micrograms per kilogram per minute
mild sedation during the introduction of the pulmonary artery cannula. Following
insertion of the cannulas, subjects rested for 30 minutes prior to measurement
of variables. Intra-arterial, central venous, pulmonary arterial, and pulmonary
capillary wedge pressures and heart rate were measured. Cardiac output (duplicate
or triplicate if duplicates differed by more than 10%, by thermodilution;
A/S3 Datex Medical Instrumentation, Tewksbury, Mass) was recorded before removal
of any blood, and 5 to 10 minutes after isovolemic removal of each 450 to
900 mL of blood into collection bags (CPDA-1 collection bags, Baxter Healthcare
Corp, Deerfield, Ill). Removal of each 450 mL of blood took approximately
10 minutes. Isovolemia was maintained by intravenous infusion of 5% human
serum albumin (Baxter Healthcare, Glendale, Calif) and/or the subject's own
platelet-rich plasma (after separation from the red blood cells of the removed
blood). These were infused simultaneously with blood removal, in quantities
approximately equal to that of the removed blood, to maintain constant central
venous and pulmonary capillary wedge pressures. At the time of cardiovascular
measurements, arterial and mixed venous blood was sampled for measurement
of pH, oxygen content, and oxyhemoglobin saturation (OSM3 Hemoximeter, Radiometer,
Copenhagen, Denmark) and arterial lactate concentration (Yellow Springs Instrument
Co, YSI No. 2700, Yellow Springs, Ohio). Stroke volume index, systemic vascular
resistance index, TO2, and O2 were calculated
by standard formulas. The subjects' pulmonary artery temperature was maintained
at 37°C by body surface warming with heated air and by warming the infused
fluids. The electrocardiogram (ECG; 5-lead in all subjects) was monitored
and a 3-channel Holter ECG in 21 of the 32 subjects (3 patients and 18 volunteers)
(Del Mar model 459, Del Mar Avionics, Irvine, Calif) was recorded continuously
from 1 hour before through the completion of the study. The frequency response
of the Holter recorder met the American Heart Association specification for
ST changes, the cutoff limit being 0.05 Hz for low frequency and 100 Hz for
high frequency. For Holter monitoring, 3 bipolar leads—CC5, modified
CM5, and ML—were used.5 Each ECG recording
on Holter tapes was scanned visually using an ECG analysis system (Del Mar
model 750). All normal QRS complexes were identified, and all abnormal QRS
complexes (eg, ventricular ectopic beats and conduction abnormalities) were
excluded from ST-segment analysis. Continuous ST-segment trends were generated
for the entire tape. All possible ischemic episodes were reviewed and verified
by an investigator who was blinded to patient identity and Hb concentration.
An ischemic episode was defined as a reversible ST-segment shift from baseline
of 0.1 mV or greater depression at J + 60 msec, or 0.2 mV or greater elevation
at the J point lasting for at least 1 minute. The time after the J point chosen
to measure ST-segment depression was adjusted to exclude T wave during tachycardia.
The relationships of measured and calculated variables to Hb concentration
were assessed by linear regression with repeated measures (JMP 3.1, SAS Institute,
Cary, NC). To test for nonlinearity of the relationship of each variable to
Hb concentration, we added a polynomial equation to the linear equation and
retested the relationship. If the fit was not improved (interaction term,P>.05), a linear relationship was accepted.
The relationship between TO2 and Hb concentration was not
linear. To find the Hb concentration at which TO2 began to decrease,
we analyzed the relationship between TO2 and Hb concentration with
a simultaneous fit of 2 linear regressions, using a population approach.6 The relationship between TO2 and Hb concentration
typically had 2 linear components: with Hb less than the inflection value
(Hbi), there was a steep relationship between Hb and TO2;
with Hb larger than Hbi, the slope of the relationship between
Hb and TO2 was shallow and either positive or negative. Our analysis
therefore determined Hbi, TO2 at Hbi [TO2(Hbi)], and the slope of each of the 2 linear portions (ie,
above [slope(Hb ≥ Hbi)] and below Hbi[slope(Hb <
Hbi)]). This was accomplished by simultaneous fitting of 2 linear
regressions that intersected at Hbi and the corresponding value
for TO2. Because the number of measurements of TO2 for
each individual ranged from 5 to 11, the number of values that could be used
to fit each of the 2 linear regressions would be small. Therefore, we applied
a method of analysis—mixed-effects modeling—in which data for
all subjects are analyzed simultaneously, accounting for interindividual differences
in such covariates as age and sex; the analysis accounts for the repeated-measures
nature of the data (NONMEM with the NONMEM program, Version V, Level 1.0L47). The model contained 4 parameters, each of which
was estimated: Hbi, TO2(Hbi), slope(Hb >
Hbi), and slope (Hb < Hbi). For the first 3 of these
parameters, interindividual variability was assumed to be log-normally distributed.
To permit the fourth parameter to be either positive or negative, interindividual
variability was modeled as normally distributed. Residual error between observed
and measured values for TO2 was assumed to have a constant variance
(ie, measurement error in TO2 did not vary with Hb). After each
population fit was complete, we used NONMEM's post hoc step to obtain Bayesian
estimates of the parameters for each individual; these Bayesian estimates
were plotted against each of age, weight, sex, and body surface area to assess
whether these covariates influenced the parameters. Goodness of fit was assessed
by inspection of the fit of the population and Bayesian regression lines to
the data for each individual and by NONMEM's objective function, − 2
log likelihood. Initially, we tested a model in which the "typical" values
of the 4 parameters were the same for all subjects. Bayesian estimates suggested
that Hbi varied with age. Permitting Hbi to vary with
age improved the fit. Bayesian estimates from this model suggested that Hbi differed between men and women. Incorporating this into the model
further improved the fit (both by visual inspection and by an improvement
in NONMEM's objective function). Bayesian estimates now suggested that slope
(Hb < Hbi) varied with age. Permitting slope (Hb < Hbi) to vary continuously with age did not improve the quality of the
fit; however, permitting slope (Hb < Hbi) to differ between
those younger and those older than 30 years of age markedly improved the fit
(again, both by visual inspection and by an improvement in NONMEM's objective
function). Bayesian estimates from this model no longer suggested a relationship
between the covariates and the parameters.
For illustrative purposes (Figure 1, Figure 2, Figure 3, Figure 4, Figure 5 and Figure 6) and to quantify the response of each variable to the reduction
in Hb concentration, data were grouped by Hb concentration increments of 10
g/L. To determine the influence of subject age and sex on the response to
acute isovolemic anemia, we performed a 2-way repeated-measures analysis of
variance. Unless otherwise specified, all data are reported as mean (SE).
Statistical significance was accepted at P≤.05.
Patients were older and of greater weight and body surface area than
the volunteers; there was no difference in initial Hb concentration or sex
proportion between the 2 groups (Table 1). The duration of experimentation was 2.4 (0.1) hours (mean [SD])
and the volume of albumin plus plasma infused was 1.1 (0.1) times the volume
of blood removed.
Right- and left-heart filling pressures (central venous and pulmonary
capillary wedge pressures) did not change with acute severe isovolemic anemia
(Figure 1, P>.05).
Decrease of Hb concentration from 131 (2) g/L (mean [SE]) to 50 (1) g/L decreased
systemic vascular resistance index by 58% (Table 2, Figure 1, P<.001), increased heart rate (Table 2, Figure 2, P<.001), stroke volume index (Table 2, Figure 2, P<.001), and cardiac index (Table 2, Figure 2, P<.001). The relationship of heart rate to Hb concentration
for each subject is depicted in Figure 3.
The increase in cardiac index compensated for the decreased arterial
oxygen content, maintaining TO2 until Hb concentration decreased
to a value (Hbi) that depended on age and sex (as determined by
population analysis). At Hb concentrations above Hbi the slope
of the relationship was × 0.669 milliliters of oxygen per kilogram per
minute (mL O2·kg−1·min−1)/grams of Hb per liter (gHb/L). The optimal model found Hbi
for males = 71.4 + 0.154·age(years) g/L and for females = 49.9 + 0.154·age(years)
g/L. Below Hbi the relationship was influenced by age: at age younger
than 30 years, slope = 12.3 (mL O2·kg−1·min−1)/(gHb/); at age 30 years or older, slope = 4.06 (mL O2·kg−1·min−1)/(gHb/L).
The TO2 at Hbi was 13.4 mL O2·kg−1·min−1.
Thereafter, compensation was incomplete, resulting in decreased TO2 (Figure 4,P<.001), and decreased mixed venous oxyhemoglobin saturation (Figure 4, P<.001).
Although TO2 decreased at the lowest Hb concentrations, O2 increased slightly (Table 2, Figure 5, P<.001).
Plasma lactate concentration did not change (Table 2, Figure 6, P=.09); arterial blood pH and base excess increased slightly
(Table 2, P<.001). There was no relationship between O2 and
TO2 in any subject.
No ST changes were observed in the online monitored ECG. Eighteen of
the 21 Holter tapes could be analyzed completely. There were 2 brief episodes
that met the criteria for significant ST changes. One episode was for 10 minutes
in a 27-year-old woman when her Hb concentration was 62 g/L when she was sitting
to urinate. Her TO2 and consumption at that Hb concentration were
normal: 16.0 and 3.9 mL O2·kg−1·min−1, respectively, and she had no cardiac symptoms. The ST changes
resolved spontaneously with reclining. Further reduction of her Hb concentration
to 49 g/L with a concomitant reduction of TO2, but not O2, did not produce ST changes. The other episode was a 0.11-mV ST depression
in a 25-year-old woman when her Hb concentration was 46 to 53 g/L. Her TO2 at that time was 10.9 to 12.1 mL O2·kg−1·min−1 and O2 was 3.9 to 4.0
unchanged from her baseline. These changes resolved at an Hb concentration
of 46 g/L when her heart rate slowed from 110 to 89 beats per minute (with
administration of esmolol, administered after the data reported here were
collected, as part of a separate protocol in a subset of these subjects).
The decrease in heart rate reduced TO2 to 8.1 mL O2·kg−1·min−1, with a O2 of
3.6 mL O2·kg−1·min−1. At the lowest Hb concentrations many subjects reported fatigue but
no other symptoms referable to decreased TO2 or hypoxia.
Sex had no influence except for women having a lesser stroke volume
index (P=.017) (and a lower Hb concentration at which
TO2 decreased; see above). Older subject age was associated with
a higher mean arterial blood pressure (P=.007) (and
a higher Hb concentration at which TO2 decreased; see above). However,
since our patients were older than the volunteers, it was not possible to
determine whether these effects related to age or population.
The major finding of this study is that acute reduction of blood Hb
concentration to 50 g/L in conscious healthy resting humans does not result
in detectable inadequate systemic TO2. The systemic markers we
used to detect consequences of inadequate TO2 (O2
and plasma lactate concentration) did not demonstrate inadequate TO2 with decreased Hb concentration. The lack of significantly increased
plasma lactate in any of the 32 subjects indicates, with a 95% assurance,
that acute reduction of Hb concentration to 50 g/L would not produce lactic
acidemia in more than 9% of the population.8
It would appear that the ST findings in the sitting subject were likely related
to body position. We were unable to determine if the 0.11-mV ST change in
the second subject was related to heart rate per se or to heart rate–induced
myocardial ischemia. Based on data from acutely anemic conscious dogs, reducing
heart rate from 110 to 89 beats per minute would decrease myocardial O2 by approximately 22%.9,10
If we accept an incidence of 1 in 18, ST changes would not be expected to
occur (with 95% assurance) in more than 16% of the population (incidence +
2 times the SE of the proportion). With an incidence of zero in 18, ST changes,
with 95% assurance, would not be expected to occur in more than 15% of the
Our data are consistent with those obtained in conscious dogs10 and baboons.11 Acute
reduction of hematocrit to 0.10 to 0.15 in conscious resting dogs and to 0.15
in conscious restrained baboons produced cardiovascular effects similar to
those we observed in humans (decreased systemic vascular resistance; increased
heart rate, stroke volume [dogs, but not baboons], and cardiac index), without
change in O2.10,11
The decreased systemic vascular resistance is, in part, related to the decreased
blood viscosity12 and, in part, to increased
cross-sectional area of the vascular bed. We accomplished intravascular dilution
with albumin and plasma so that altered viscosity in our subjects resulted
solely from the decreased red cell concentration.
In our subjects, increased heart rate contributed approximately 75%
and increased stroke volume approximately 25% of the increased cardiac output.
These values are similar to those of 69% and 31%, respectively, in acutely
anemic conscious dogs.10 Inadequate TO2 results in an inability to sustain O2. Our subjects
actually increased their O2 by approximately 14% at an Hb
concentration of 50 g/L. This is likely due to the increased heart rate and/or
sympathetic stimulation. If the relationship between myocardial O2 and heart rate during acute anemia is similar to that of the dog,9,10 increased myocardial O2 accounted for approximately one fourth of our measured increased systemic O2. Our result is consistent with the finding by Cain13
of a small increase in the O2 of anesthetized dogs, at a
hematocrit just above that which decreased O2 (hematocrit
approximately 0.10). We found that TO2 was maintained at Hb concentrations
of at least 65 g/L for a 20-year-old woman and 92 g/L for a 20-year-old man.
These values increased with increasing subject age. Our finding differs from
that in conscious dogs of a linear decrease in TO2 with decreasing
Hb concentration.10 The decreased TO2, however, does not imply inadequate delivery of oxygen to tissues.
Cain defined the "critical" TO2 as that which is inadequate
to prevent a decrease in O2.13
The critical systemic value varies among species: approximately 10 mL O2·kg−1·min−1 in anesthetized
dogs with mechanically ventilated lungs,13
and 23 mL O2·kg−1·min−1 in anesthetized rats.14 In anesthetized
dogs, systemic TO2 becomes critical at an Hb concentration of 30
to 50 g/L.15 In dogs, the critical myocardial
TO2 occurs at approximately the same Hb concentration as does the
systemic TO2.16 This adds support
to our interpretation that the ST change that occurred in the one volunteer,
in the absence of evidence of inadequate systemic TO2, was likely
related to heart rate per se and not myocardial ischemia. Reducing hematocrit
to 0.17 in conscious dogs did not decrease splanchnic O2
or hepatic function,17 and reducing hematocrit
to 0.15 in anesthetized pigs did not alter systemic or hepatic O2 or hepatic lactate uptake.18 Oxygen
delivery of 16 mL O2·kg−1 ·min−1 at a hematocrit of 0.10 to 0.15 in conscious resting dogs did
not produce evidence of inadequate TO2,10
nor did delivery of 11 mL O2·kg−1·min−1 at a hematocrit of 0.15 in conscious restrained baboons.11
The "critical" value of TO2 in humans has not been determined.
An evaluation of the reports of clinical experiences with Jehovah's Witnesses
who refused transfusion suggested that an Hb concentration of 50 g/L or greater
is not associated with increased mortality, but could not assess the level
of critical TO2.19 van Woerkens
et al20 reported a critical value of 4.9 mL
an anesthetized 84-year-old man with pharmacologically induced neuromuscular
blockade and mechanically ventilated lungs. The critical value for Hb or TO2 has not been determined in any conscious species at rest. Reducing O2 by anesthesia, neuromuscular blockade, and mechanical ventilation
of the lungs would likely decrease the critical value for TO2 below
that of the normal conscious state. Anesthetics decrease systemic O2, myocardial contractility, and thereby myocardial O2.21 Anesthetic-induced cortical electrical silence reduces
cerebral O2 by approximately half.22
Neuromuscular blockade reduces muscle O2 by eliminating movement.
Mechanical ventilation of the lungs decreases the O2 of the
muscles of respiration, but also decreases venous return, and thereby cardiac
output. Additionally, many anesthetized humans are mildly hypothermic, further
decreasing O2. Most anesthetics also decrease cardiac output.21 Thus, data from anesthetized animals or humans cannot
be extrapolated to the conscious state because of alteration of both O2 and TO2. Similarly, our data cannot be extrapolated to
conditions other than rest. In dogs, exercise decreases the ability to compensate
for acute anemia: systemic O2 decreases10
and the coronary arteries dilate maximally9
at higher Hb concentrations than during rest.
Our effort to determine the critical level of TO2 in conscious
resting humans failed. Mean TO2 was 10.7 (0.4) mL O2·kg−1·min−1 at an Hb level of 50 (1) g/L.
Fourteen of our 32 subjects had TO2 of less than 10 mL O2·kg−1· min−1; the lowest
TO2 achieved was 6.5 mL O2·kg−1·min−1. Nevertheless, we did not find evidence of systemic acidosis
in any subject. The slight increases of arterial blood pH and base excess
were likely a result of the infused citrate anticoagulant (contained within
the plasma). The electrocardiographic ST changes in 1 subject could have been
related to heart rate or could have represented transient myocardial ischemia.
Although we used a standard measure of myocardial ischemia that provides continuous
measurement and printed availability for validation,23
it remains possible that some episodes of ischemia might not have been detected.24
"Guidelines," "standards," and "practice parameters" have suggested
that patients without cardiovascular or pulmonary disease need not be given
red blood cell transfusion unless their Hb concentration is less than 70 g/L1 or 60 g/L.3 The development of these documents and the
advice they contain has been hampered by limited human data on which the authors
and committees could rely. Our data support these recommendations and suggest
that in healthy resting patients not taking drugs with cardiovascular actions,
even lower Hb concentrations are tolerated without apparent adverse metabolic
sequelae. We were unable to determine the critical Hb value in our healthy
subjects, and thus cannot provide a definitive Hb concentration likely to
require red blood cell transfusion. However, the human critical value for
TO2 is less than 10 mL O2·kg−1·
min−1, a level reached in normal humans at an Hb concentration
of 50 g/L. Our subjects were studied supine, at rest. Thus, our results should
not be extrapolated to circumstances of increased O2 (eg,
activity), decreased ability to increase cardiac output (eg, cardiac disease,
medications), or decreased ability to increase specific organ or tissue blood
flow (eg, arterial stenosis).