Fallibility of the Intravenous Glucose Tolerance Test as a Measure of Endogenous Glucose Turnover Robert R. Wolfe,
John R. Allsop, and John F. Burke
We have used hepatectomized, nephrectomized dogs receiving a constant infusion of unlabeled glucose as well as conscious, unrestrained guinea pigs in order to investigate the calculation of basal glucose kinetics from intravenous glucose tolerance tests (IVGTT). In the dogs, we were not able to determine the known rate of appearance (R,) or disappearance (R,,) of glucose within 50% of the actual value by means of IVGTT. In the guinea pigs, we found that R, calculated from IVGTT was 250% higher than R, determined by
means of the validated technique of the primed-constant infusion of 6-‘H-glucose in tracer quantities. When live Escherichia co/i were infused into the guinea pigs, the isotope-tracer technique revealed a 100% increase in R, yet R, appeared to be decreased by 80% when calculared by means of IVGTT. We concluded that basal glucose kinetics cannot be determined reliably from IVGTT, and that in certain pathologic conditions the direction of change in R, and R, from the basal state may be incorrectly predicted.
0
VER THE PAST 25 yr many attempts have been made to quantitatively assess basal glucose kinetics from intravenous glucose tolerance tests (IVGTT). Most commonly, the rate of decay in plasma glucose concentration after the glucose injection has been used to calculate the rate constant (k value). The slope of the glucose decay curve (plotted on a semilogarithmic scale) has been extrapolated to the time of glucose injection (t=O) in order to calculate the glucose pool size. In a dynamic steady state the rate of uptake of glucose equals the rate of production and this can be calculated from the pool size and the k value. A formula for such a calculation of the rate of appearance of glucose (R,) was described by Hlad et al.’ in 1956, and the same technique has been used recently by Wilmore et al.2J to investigate alterations in glucose kinetics following burn injury and sepsis. Despite the years that have passed since the original description of the technique by Hlad et al.,’ a rigorous attempt to validate it has not been made. This is disturbing since in their original description Hlad et al. found that their values varied greatly between subjects and were not reproducible from day to day in the same individual.’ They also obtained significantly different results when a constant infusion of glucose was used rather than a bolus injection, Finally, their conclusion that “no evidence was obtained to support the concept that intravenous hyperglycemia stimulates the secretion of extra insulin” raises serious questions as to the reliability of their technique. We therefore felt that it was necessary to evaluate the reliability of basal glucose kinetic data derived from glucose tolerance tests. Although the calculations described by
From the Department General Hospital:
OJ Surgery.
Harvard
Medical
School;
and Surgical Research, Shriners Bums Institute.
Received Jor publication Supported by NIH
Surgical
Services,
April 13. 1977.
Grants GM 070351
and GM 021700.
Reprint requests should be addressed to Robert R. Wolfe, 51 Blossom Street, 0 1978 by Grune & Stratton,
Metabolism,
Massachuselts
Boston, Mass.
Boston, Mass. 02114.
Inc. 0026-0495/78/2702-0010$02.00/0
Vol. 27, No. 2 (February),
1978
217
218
WOLFE,
ALLSOP,
AND BURKE
Hlad et al.’ are not the only method of analysis, they are representative of the principles and assumptions involved in any mathematic analysis of the IVGTT. The investigation of the Hlad technique is therefore of relevance to other analyses4 that are but variations on the theme of a pool size and a rate constant. It is self-evident that if the correct rate of production and uptake of glucose cannot be determined from the li value and the pool size, then the li value and/or the pool size must also be incorrect. Since the pool size is derived from the k value by extrapolation, it is likely that both factors would be in error simultaneously. An inability to predict R, from a glucose tolerance test would therefore raise serious questions as to the physiologic significance of the k value. We have used two experimental approaches to investigate the validity of R, calculated from pool size and k values derived from an IVGTT. In one. we removed endogenous sources of glucose production in the dog by hepatectomy and bilateral nephrectomy and infused glucose at a known rate. We then compared the R, predicted by the technique described by Hlad et al.’ to the known infusion rate. In the second approach, we used intact animals, both normal and septic, to compare the values of R, obtained from IVGTT with those obtained by means of the primed-constant infusion of 6-3 H-glucose. We have previously demonstrated that the primed-constant infusion of 6-3H-glucose into a dog deprived of endogenous glucose production can measure a known rate of infusion of unlabeled glucose with a coefficient of variation of 3.S”,,.5 The results of our study demonstrate that the technique described by Hlad and associates’ cannot even approximately predict the R,. Furthermore, when their technique is applied during pathologic states such as sepsis, a dramatic decrease in R, may be calculated to have occurred when, in actual fact, R, is increased. MATERIALS We first attempted to validate was known. performed.
AND
the Hlad
In three anesthetized
METHODS
et al. technique’
and this was followed
by total hepatectomy
and associates.6
Glucose
(using a Harvard
infusion pump) in order to maintain
proximately
constant
III a system in which
dogs* that had been fasted for 36 hr. bilateral
was then infused
level. After
using the technique
into the dogs at the exact
2.05
had been maintained
for
2 hr. which was sufficient time for a dynamic
steady state to be well established.
cose (300
over
mg/kg)
was injected
over the next 2 hr and analyzed (Beckman
Instruments,
intravenously
The
proportionality
mg/kg/min)
Fullerton,
Calif.).
In accord
for the duration
constant
(X/min).
were then calculated
mg/kg/min
with
the assumption
by a bolus injection
a bolus of glu-
samples
on a Beckman
at an ap-
approximately were
drawn
glucose analyzer
made by Hlad
of glucose.
et al.’
we continued
the
R,
(in
of the experiment. the
as described
the decay of plasma glucose concentration cose on a semilogarithmic
I7 arterial
for plasma glucose concentration
that the basal glucose kinetics are not alfected infusion of glucose unaltered
30 sec. and
glucose by Hlad
space et al.’
versus minutes
(I ~ in This
ml/kg).
procedure
after the injection
scale. The log of glucose concentration
declined
and
the
involved
plotting
of the bolus of gluin a linear
manner
between IO and 60 min, and this segment of the glucose decay curve was used to determine value. The straight
*In
conducting
Facilities
line fitted to the log glucose concentration
this research,
the investigators
and Care as promulgated
sources, National
Academy
was
described by Markowitl rate of
the plasma glucose concentration
the glucose infuslon
R,
the actual
nephrectomy
adhered
by the Committee
of Sciences- National
always
to the
Guide /or
for Laboratory
Research Council.
passed within
Animal
+2”,,
the k of the
Laboraror~~ Anirnul Facilities
and Re-
FALLIBILITY
219
OF IVGTT
actual measured value. The line was extrapolated to give the theoretic glucose concentration when t = 0 (CO): the glucose space (V) was calculated by dividing the quantity of gluyse administered in the bolus by (Ca - C,,), where C,, is the “experimental fasting level,” or the concentration of glucose when the glucose tolerance curve levels off (for an example, see Fig. I). Total body glucose (G,, in mg/kg) was obtained by multiplying V x C,,. The roportionality constant k was calculated from the following equation: C,, - Ccs = (Cc - C,,)e, where C, is the arterial glucose concentration at r = x. The rate of uptake of glucose (Rd) and the R,, which are equal when the animals are in a dynamic steady state (as they were in our experiment), was calculated by multiplying k x G,. The second aspect of our investigation was to compare results obtained in intact animals by means of the Hlad et al. technique’ with those obtained by means of the primed-constant infusion of 6-3H-glucose (New England Nuclear, Boston, Mass.). In order to assess the applicability of the Hlad et al. technique to the study of glucose kinetics during stress or trauma. we performed studies in both normal guinea pigs and guinea pigs infused with live Escherichia co/i. The guinea pigs had chronic catheters (PE 50) surgically implanted (halothane anesthesia) in the carotid artery and jugular vein 48 hr before the experiment. The animals were fasted from the time of surgery until the start of the experiment. but were provided with water ad libitum. All experiments were performed in unanesthetized. unrestrained animals. In one group of animals (n = 4). we compared the values obtained for R, by means of the IVGTT and by means of the primed-constant infusion of 6-3H-glucose both before and 4 hr after the infusion of approximately IO lo live E. coli bacteria. The procedure for determining R, by means of the primed-constant infusion of radiolabeled glucose has been in use for many years,’ and we have validated this technique experimentally in the hepatectomized, nephrectomized dog preparation used in this study.’ The selection of 6-3H-glucose as an appropriate tracer molecule has been dealt with by Altzuler et a1.s Between 14 and 2 hr after the start of the 6-3H-glucose infusion. three blood samples (0.2 ml) were drawn. and the plasma glucose specific radioactivity was determined.g Since the animals were in an isotopic steady state by that time, R, was calculated by dividing the isotope infusion rate by the mean plasma glucose specihc activity.9 Then 500 mg/kg of unlabeled glucose was administered intravenously as a bolus; arterial blood samples (0.07 ml) were drawn at 5. 7.5, IO. 15, 20, and 30 min after the unlabeled glucose infusion and analyzed for glucose concentration. Three hours were then allowed for the animal’s blood glucose to stabilize at the original level, after which the bolus of live E. coii bacteria was injected through the arterial catheter without interruption of the 6-3H-glucose infusion. At 3% and 4 hr after the E. co/i injection. two blood samples were drawn to determine R, by means of the radiolabeled tracer technique.’ Another bolus of unlabeled glucose (500 mg/kg) was then injected intravenously, with samples again being drawn at 5, IO, 15. 20. and 30 min in order to calculate R,. k, and V according to Hlad et al.’ The time course of the elfects of E. co/i injection on R, determined by means of the primedconstant infusion of 6-‘H-glucose was studied in six guinea pigs. After isotopic steady state had been attained and basal R, determined. these animals were infused with approximately IO’” live E. co/i. Arterial samples were then drawn every hour and R, and Rd were calculated as described before.9 The suppressibility of endogenous glucose production by an exogenous infusion of unlabeled glucose was evaluated in a final group (n = 5). The same basic protocol was used as described above. except that a constant infusion of unlabeled glucose was started and maintained at the rate of IO mg/kg/min between the second and fourth hour postinfusion of E. coli. An isotopic steady state was achieved by I$ hr after the start of the unlabeled infusion, and endogenous R, was determined during the next 30 min of the infusion by means of the radiolabeled tracer technique.g We had previously found this rate and duration of exogenous glucose infusion to lower IO endogenous R, 67”,, in normal guinea pigs. RESULTS
The data from the glucose tolerance three hepatectomized, nephrectomized
test (IVGTT) administered to one of the dogs are presented in Fig. 1. When the
WOLFE,
ALLSOP,
AND
BURKE
oo( I-
5a )3ccP
ICKI-
5c )L
3c I-
!B, I80
205
230
255
280
Fig. 1. (A) Plasma glucose concentration in naphrectomized, hepatectomired dog (No. 3) receiving Q constant infusion of glucose (2.05 mg/kg/ min) from t = 0 to t = 275 min. The IVGTT (300 mg/kg) was started at t = 155 min. (B) IVGTT data from dog 3 plotted on semilogarithmic fasting level” of glucose. scale. C,. . “experimental
data were plotted on a semilogarithmic scale (Fig. I), Co was found to be 330 mg/dl. C,, was 57 mg/dl, although the plasma glucose concentration was 87 mg/dl before the administration of the IVGTT. For this dog, k was calculated to be O.O278/min, V was 110 ml/kg, and R, was 3.06 mg/kg/min. The calculated R, was therefore 49”,, higher than the actual R, of 2.05 mg/kg/min. The values obtained from the other dogs were similar in most respects (Table I), although the calculated value for R, was overestimated to an even greater extent in the other dogs. In order to calculate the basal glucose kinetics from an IVGTT administered to guinea pigs. we curve-fitted a line to the mean blood glucose concentration for the group obtained at each time interval after the bolus of glucose. This modification of the technique of Hlad et al.’ has been used frequently when the IVGTT have been administered to small animals.” The resulting curves obtained before and 4 hr after the infusion of live E. co/i are presented in Figs. 2 and 3. The values calculated on the basis of these curves are presented in Table 2. The most striking finding was the impairment of “glucose tolerance” after the E. coli infusion, reflected by a decrease in k from O.O54/min before the E. Table 1, Glucose Kinetics Calculated KIW,Vn Animal
From IVGTT in Hepatectomired,
Proportionality Constant
R,*
(min-‘)
(k)
Glucose Size
NO.
(w/kg/minj
1
2.05
.0317
117
2
2.05
.0290
3
2.05
.0278
*R,: rate of appearance
of glucose.
W/kg)
Nephrectomized
Pool
Dogs
Error Calculated f%
of
Calcvloted R,
(%I
3.18
+55
125
3.66
+79
110
3.06
t49
221
FALLlBlllTY OF IVGTT
r
1oot
Mean plasma gluFig. 2. mse concentration following injection of 500 mg/kg gluinto normal, fasting case guinea pigs (N = 4).
I Oi
coli infusion to O.O068/min at 4 hr after the infusion. These changes in k value corresponded to a reduction in R, from 17.36 mg/kg/min before the E. coli infusion to 3.59 mg/kg/min after E. cob. In contrast, R, determined by the primed-constant infusion of 6-3H-glucose (Ra3H) revealed a mean value of 4.81 mg/kg/min before E. coli and increased to 9.64 mg/kg/min 4 hr after the E. cofi infusion (Table 3). The results of the time course study of R,-‘H in six guinea pigs infused with E. coli are presented in Fig. 4. During the first 6 hr after the E. co/i infusion, both Ra3H and R:H were always significantly elevated (according to single
Fig. 3. Mean plasma glucose concentration following injection of 500 mg/kg glucose into fasting guinea pigs (N = 4) that had been infused with 1O’O live C. coli 4 hr earlier.
I
I/I
0
5
10
15
20
25
30
MII’JLJ TES
Table 2. Glucose Kinetics Calculated
From IVGTT in Guinea
Proportionality
GllJCOSe
Constant (k)
Pool Size (V)
(mine’)
(44
Pigs GlWXe
Turnover (R,) (m/Wmin)
Control
0.054
255
17.36
4 hr post E. co/i
0.0068
285
3.59
Values
represent
calculations
from
pooled
data
from
four
animals.
WOLFE,
222
Table
3.
Individual
Rates
Means
of Appearance
of the
of Glucose
Primed-Constant
(R,)
Infusion
in Guinea
ALLSOP,
Pigs
AND
Determined
BURKE
by
of 6-3H-Glucose R, (w/kg/min)
Animal
4 hr COiTtd
No.
Studies
were
Post E. Co11
1
5.42
2
3.69
7.03
3
4.43
10.61
4
5.68
11.09
Mean
4.81
9.44
done
in the some animals
in which
IVGTT
9.81
were
administered
(Table
2).
factor analysis of variance and Dunnet’s t test, p < 0.05). During the first 4 hr, the plasma glucose concentration rose because Ra3H exceeded Rd3H. Between 4 and 6 hr after E. coli infusion (the time corresponding to when the IVGTT were given), Re3H remained significantly elevated but the plasma glucose concentration fell slightly, indicating an even greater increase in Rd3H. When glucose was infused at the rate of 10 mg/kg/min for 2 hr into animals given live E. cofi, there was no suppression of endogenous R,-‘H whatsoever and the plasma glucose concentration rose to a mean value in excess of 500 mg/dl (Fig. 5). We have previously shown that the same procedure suppressed endogenous Ra3H 67”,, in normal guinea pigs.”
r
i
Fig.
4.
Rate
(R,),
ance
(R.j),
cose
concentration
the
rate
of
ance
and
infusion
of
lOlo
live
were
determined
the
of
E.
appear-
disappear-
plasma
approximately
co/i.
R, by
and means
primed-constant
of 6-3H-glucose. of individual
glu-
following
data
Rd of
infusion
Mean
f
(N
= 6).
SEM
FALLIBILITY
223
OF IVGTT
Fig. 5. Plasma glucose conendogenout centration and rate of appearonce of glucose (R,) in experimental sepsis before and during (I Z-hr infusion of glucose (10 mg/kg/ min) as compared to control value. R, was determined by means of the primed-constant 6-3H-glucose. infusion of Mean f SEM of individual data (N = 5).
L
100
0
POST E.coli
POST E,coli GLUCOSE
DISCUSSION
A number of assumptions must be made in order to calculate the rate of production of glucose (R,) from a glucose tolerance test. Fundamental among these is the belief that it is valid to curve-fit data from a glucose tolerance test and that each term in the mathematic description of that curve has a physical correlate such as the glucose pool size or the rate of tissue uptake of glucose. Other assumptions made by Hlad et al.’ are (1) that there is a single glucose space in which a bolus of injected glucose is uniformly mixed within 10 mitt, (2) that the tissue uptake of glucose is linear from 100 to 600 mg/dl. (3) that all disappearance of glucose from the sampled glucose pool is attributable to the tissue uptake of glucose, and (4) that an intravenous bolus infusion of glucose does not disturb the basal glucose kinetics. Other mathematic analyses of the IVGTT share many of these assumptions, although some writers assume that the glucose bolus reduces endogenous glucose production to zero,4 so that the k value is taken to represent tissue uptake only rather than the net balance between tissue uptake and continued production. This alternative method of analysis4 is addressed in the discussion of the guinea pig data. The prediction of R, from IVGTT in hepatectomized, nephrectomized dogs receiving a constant infusion of unlabeled glucose would be expected to be more reliable than doing so in normal dogs because, contrary to physiologic responses, ‘* our dog preparation complied with two of the assumptions of Hlad et al:’ first, the basal R, was not affected by the bolus injection of glucose, and second, in the nephrectomized dog there are no nonlinear losses of glucose in urine due to tubular reabsorption capacity being exceeded during an IVGTT.13 The inability to predict R, accurately in a system “loaded” to comply with assumptions is therefore striking evidence that the IVGTT and the equation of Hlad and associates cannot reliably measure R,. It should be emphasized that
224
WOLFE,
ALLSOP,
AND BURKE
in six studies using the same animal model, the primed-constant infusion of 6-3H-glucose measured R, with a coefficient of variation of 3.5”,, about the true value.5 The question arises as to why the Hlad equation fails to predict R, in a model designed to comply with initial assumptions. The work of Steele and associates14 as well as our own’ has established that the volume of distribution of glucose consists of at least two theoretic pools. Steelei conceptualized these pools as consisting of a fast-mixing compartment, which our isotope studies indicate corresponds roughly to the plasma volume.5 and a slow-mixing pool equal to the remainder of the glucose space, which is presumably interstitial fluid. Thus, the rapid decay of glucose concentration during the first IO min after the bolus injection of glucose probably represents mixing in the plasma compartment. The rate of decay in plasma glucose concentration (li/min) after the completion of mixing in the plasma compartment is then determined by the balance between the rate of diffusion of glucose from the plasma to interstitial fluid and the rate of continuing glucose production. The rate of diffusion from plasma to interstitial fluid following an IVGTT will depend on the concentration gradient across the capillary wall and any further concentration gradients within the interstitial fluid compartment. The rate of tissue uptake of glucose is one factor influencing the interstitial fluid glucose concentration but is not the sole factor determining the gradient between plasma and interstitial fluid. For example, the percentage increment in plasma glucose concentration and hence diffusion gradient will depend on the size of the glucose bolus relative to the plasma glucose concentration at the time it is injected. On this basis a standard glucose bolus in a hyperglycemic situation would establish a smaller percent increment in plasma glucose. a smaller diffusion gradient between plasma and interstitial fluid, and hence a smaller k value. This argument agrees well with the relationship between the k value and total glucose pool size observed by Wilmore et al.3 It also explains the observation of CahillI that when a glucose tolerance test is repeated in a diabetic who has no insulin response, the k value is inversely proportional to the blood glucose concentration at the start of each test. The failure to correctly calculate R, by the Hlad technique in our dogs supports the argument that factors other than rate of tissue uptake significantly influence the k value even when R, is kept constant. In intact animals the application of IVGTT to measure R, is further compounded by the unknown extent of suppression of endogenous glucose production by the glucose bolus. The evidence in this paper and previously published results” indicate that exogenous glucose infused at IO mg/kg/min depresses endogenous production approximately 67”,, in normal guinea pigs and not at all in guinea pigs given live E. cofi. The suppressibility of gluconeogenesis after of R, from IVGTT in burn injury is again different. lo The 250”,, overestimate normal guinea pigs can be attributed in part to the assumption of Hlad et al. that endogenous R, proceeds uninterrupted after the glucose bolus. However. if we had not made that assumption and therefore did not use the C,, value in the calculations, the R, would still have been overestimated by 67”,,. On the other hand, the failure to use C,, in the calculations for the septic animals
FALLlBlllTY
OF IVGTT
225
magnified the already large error encountered when the C,, value was used. The lack of suppressibility of endogenous R, in the septic animals probably accounted for the desirability of using the C,, value in that situation. Unfortunately, there is no way to know a priori whether the use of the C,, value will improve or decrease the accuracy of the calculated R,, and in any event substantial errors persist regardless of its inclusion or deletion. The errors in calculated R, may also be partly due to the unfounded assumption that mathematic terms, such as the k value, that describe successive data points have physical correlates. Berlin and associates” have pointed out that a single exponential curve can approximate a process that is in fact a two-compartment constant coefficient system containing no less than four different turnover rates. The danger of assigning physical characteristics to the terms used in the mathematic descriptions of the data is increased by the fact that with IVGTT the data points frequently lie further from the best fit monoexponential curve that can be explained by experimental error. Thus, in intact animals, the combination of an undefined diffusion gradient, a variable suppressibility of endogenous R,, and difficulties in equating mathematic terms with physiologic events can result in an R, predicted from an IVGTT that has no relationship to the actual R,. For example, using the validated isotopic technique’ R, was elevated loo”,, after E. coli. yet when calculated by the IVGTT method R, was depressed 80”,, after E. coli. A final comment is required about the use of pool size and k value as isolated indices of glucose kinetics in vivo. Fortuitously, in the calculation of R, the error in k is offset to some extent by the error in glucose pool size, which is determined by extrapolation of k. It would therefore appear that the glucose pool size and k value considered separately are even less reliable indices of glucose metabolism than is the R, calculated from them. This is exemplified by the results presented in Fig. 1. With only one exception, the data points from 10 to 70 min after the glucose injection fell on a single exponential curve, yet when that curve was extrapolated to t = 0 in order to calculate the total glucose space (V), a value was obtained that was less than one-half the value originally established by Wick et al.” as the glucose space of hepatectomized, nephrectomized dogs. The unreliability of the k value is particularly noteworthy, since the k value is widely believed to be synonymous with the rate of tissue uptake of glucose. The flat curves we observed in septic guinea pigs (Fig. 3) are almost identical to those that have been reported following several forms of trauma.““’ Such “diabetic” IVGTT curves have classically been used to diagnose “insulin resistance”-an inability of tissues to take up glucose after trauma.” However, the validated radioisotope data reveal that the tissues can actually be taking up glucose at twice the normal rate at a time when the k value is decreased markedly. The conflict of these findings with the traditional interpretation of k results from the fact that the rate of tissue uptake of glucose is only one factor contributing to the k value. The extent of suppression of gluconeogenesis as well as the undefined diffusion gradients within the glucose pools can dominate the shape of the IVGTT, and consequently the k value should not be used as an index of glucose uptake.
226
FALLIBILITY
OF IVGTT
REFERENCES I.
Hlad CJ Jr, Elrich
on the
kinetics
Invest 35: I I39 2. Wilmore
I 149,
1956
DW.
Mason
in
glucose
Jr: Alterations thermal
H, Witten
of glucose
TA: Studies
utilization.
3. Wilmore
DW.
Mason
glucose
flow
with gram-negative 143:720
Jr. Pruitt
kinetics
BA
following
AD
Jr. Pruitt
in burned
the intravenous 23:103-l
BA
Obstet
D: Glucose
assimila-
evaluation
of
glucose tolerance
test. Clin
Sci
reliability
Wolfe
RR,
Burke
Jl-:
of rates of glucose appearance
calculated
from constant
tracer
The
in vivo
infusions.
Bio-
them J (in press) 6. Markowitz
J. Yates
WM.
A simple one stage technique in thedog.
Burrows
J Lab Clin Med IX:1271
fect of insulin on utilization glucose.
Am
and production
by a new tracer
01
1X9:43 50.
14. Steele R.
Glucose
Barkai
turnover
with various
tn the
burn
guinea
pigs.
dog
C. et al: obtained
glucose.
Am
Glucose Phyrml
Miller
injury Surg
HI. on
dtabetic
method.
dogs
Metabolism
of the Kidney
studies
in
and
Year Book. 1971
hepatic
R: Does
sugar output’!
evrscerated
dogs.
a
14C
Am
J
on pools. Fed Proc
679. I Y6-t Physiology
Diabetes 20:7X5 17. Berlin
N.
Berman
The application Med 6X:423
of insulin tn man.
799. I97 I M.
Berk
PO,
ofmulticompartmental
of cltnical
medictne.
et al:
analv\ts Ann
Intern
436. 196X AH.
Drury
DR.
Makay
EM:
Glu-
cose space of the body. .Am J Phystol 163:224 27x. 1950
Elaht D. et al: Efglucose
Gynecol
Carbo-
infection.
197:60 62. 1959
IX. Wick J
1667. 1975
RR.
fect of 364, I977
values
Bjerknes
species of labeled
Phvsiol229:1662 9. Wolfe
A.
233:E80-
Jr:
Btshop JS. Levine
glucose load inhibit
to problems N.
burn
and re-
Jr: Glucoregula-
and
Physiology
16. Cahill GF.
1957 8. Altzuler
of
372, 1971
Body Fluids. Chicago,
I
AD
G
in normal
13. Pitts RF-
22:67
127X. 1933
J Physiol
Mason
15. Steele R: Retlections
WH.
for hepatectomy
7. Wall JS. Steele R. de Bodo RC. et al: Efcirculating
tlfect
oxidation
in Pseudomonas
JS. Hetenyt
responses
20:360
A critical
JR,
Jl-:
Biomed Rea 2:C I 67. 196X
Comput
tory
Kl-.
metabolism
12. Cowan
13. 1962
5. Allsop
LaNoue
recorded
V. Marrack
Burke
E85, 1977 II.
patients
sepsis. Surg Gynecol
tion in hyperinsulinism.
RR,
on glucose turnover.
hydrate
83. 1975
724, 1976
4. Marks
IO. Wolfe trauma
cycling in guinea pigs. Am J Physiol AD
injury. Surg Forum 25:81
Jr: Impaired
J Clin
turnover
Obstet
In
144:359--
IY.
Frayn
KN.
sulin secretion the rat in viva. 1975
Eti’ects ol burn injury on in-
and on sensttivity Eur J Clin
to insultn
Invest
5:33l
in
337.
John R. Allsop, and John F. Burke
We have used hepatectomized, nephrectomized dogs receiving a constant infusion of unlabeled glucose as well as conscious, unrestrained guinea pigs in order to investigate the calculation of basal glucose kinetics from intravenous glucose tolerance tests (IVGTT). In the dogs, we were not able to determine the known rate of appearance (R,) or disappearance (R,,) of glucose within 50% of the actual value by means of IVGTT. In the guinea pigs, we found that R, calculated from IVGTT was 250% higher than R, determined by
means of the validated technique of the primed-constant infusion of 6-‘H-glucose in tracer quantities. When live Escherichia co/i were infused into the guinea pigs, the isotope-tracer technique revealed a 100% increase in R, yet R, appeared to be decreased by 80% when calculared by means of IVGTT. We concluded that basal glucose kinetics cannot be determined reliably from IVGTT, and that in certain pathologic conditions the direction of change in R, and R, from the basal state may be incorrectly predicted.
0
VER THE PAST 25 yr many attempts have been made to quantitatively assess basal glucose kinetics from intravenous glucose tolerance tests (IVGTT). Most commonly, the rate of decay in plasma glucose concentration after the glucose injection has been used to calculate the rate constant (k value). The slope of the glucose decay curve (plotted on a semilogarithmic scale) has been extrapolated to the time of glucose injection (t=O) in order to calculate the glucose pool size. In a dynamic steady state the rate of uptake of glucose equals the rate of production and this can be calculated from the pool size and the k value. A formula for such a calculation of the rate of appearance of glucose (R,) was described by Hlad et al.’ in 1956, and the same technique has been used recently by Wilmore et al.2J to investigate alterations in glucose kinetics following burn injury and sepsis. Despite the years that have passed since the original description of the technique by Hlad et al.,’ a rigorous attempt to validate it has not been made. This is disturbing since in their original description Hlad et al. found that their values varied greatly between subjects and were not reproducible from day to day in the same individual.’ They also obtained significantly different results when a constant infusion of glucose was used rather than a bolus injection, Finally, their conclusion that “no evidence was obtained to support the concept that intravenous hyperglycemia stimulates the secretion of extra insulin” raises serious questions as to the reliability of their technique. We therefore felt that it was necessary to evaluate the reliability of basal glucose kinetic data derived from glucose tolerance tests. Although the calculations described by
From the Department General Hospital:
OJ Surgery.
Harvard
Medical
School;
and Surgical Research, Shriners Bums Institute.
Received Jor publication Supported by NIH
Surgical
Services,
April 13. 1977.
Grants GM 070351
and GM 021700.
Reprint requests should be addressed to Robert R. Wolfe, 51 Blossom Street, 0 1978 by Grune & Stratton,
Metabolism,
Massachuselts
Boston, Mass.
Boston, Mass. 02114.
Inc. 0026-0495/78/2702-0010$02.00/0
Vol. 27, No. 2 (February),
1978
217
218
WOLFE,
ALLSOP,
AND BURKE
Hlad et al.’ are not the only method of analysis, they are representative of the principles and assumptions involved in any mathematic analysis of the IVGTT. The investigation of the Hlad technique is therefore of relevance to other analyses4 that are but variations on the theme of a pool size and a rate constant. It is self-evident that if the correct rate of production and uptake of glucose cannot be determined from the li value and the pool size, then the li value and/or the pool size must also be incorrect. Since the pool size is derived from the k value by extrapolation, it is likely that both factors would be in error simultaneously. An inability to predict R, from a glucose tolerance test would therefore raise serious questions as to the physiologic significance of the k value. We have used two experimental approaches to investigate the validity of R, calculated from pool size and k values derived from an IVGTT. In one. we removed endogenous sources of glucose production in the dog by hepatectomy and bilateral nephrectomy and infused glucose at a known rate. We then compared the R, predicted by the technique described by Hlad et al.’ to the known infusion rate. In the second approach, we used intact animals, both normal and septic, to compare the values of R, obtained from IVGTT with those obtained by means of the primed-constant infusion of 6-3 H-glucose. We have previously demonstrated that the primed-constant infusion of 6-3H-glucose into a dog deprived of endogenous glucose production can measure a known rate of infusion of unlabeled glucose with a coefficient of variation of 3.S”,,.5 The results of our study demonstrate that the technique described by Hlad and associates’ cannot even approximately predict the R,. Furthermore, when their technique is applied during pathologic states such as sepsis, a dramatic decrease in R, may be calculated to have occurred when, in actual fact, R, is increased. MATERIALS We first attempted to validate was known. performed.
AND
the Hlad
In three anesthetized
METHODS
et al. technique’
and this was followed
by total hepatectomy
and associates.6
Glucose
(using a Harvard
infusion pump) in order to maintain
proximately
constant
III a system in which
dogs* that had been fasted for 36 hr. bilateral
was then infused
level. After
using the technique
into the dogs at the exact
2.05
had been maintained
for
2 hr. which was sufficient time for a dynamic
steady state to be well established.
cose (300
over
mg/kg)
was injected
over the next 2 hr and analyzed (Beckman
Instruments,
intravenously
The
proportionality
mg/kg/min)
Fullerton,
Calif.).
In accord
for the duration
constant
(X/min).
were then calculated
mg/kg/min
with
the assumption
by a bolus injection
a bolus of glu-
samples
on a Beckman
at an ap-
approximately were
drawn
glucose analyzer
made by Hlad
of glucose.
et al.’
we continued
the
R,
(in
of the experiment. the
as described
the decay of plasma glucose concentration cose on a semilogarithmic
I7 arterial
for plasma glucose concentration
that the basal glucose kinetics are not alfected infusion of glucose unaltered
30 sec. and
glucose by Hlad
space et al.’
versus minutes
(I ~ in This
ml/kg).
procedure
after the injection
scale. The log of glucose concentration
declined
and
the
involved
plotting
of the bolus of gluin a linear
manner
between IO and 60 min, and this segment of the glucose decay curve was used to determine value. The straight
*In
conducting
Facilities
line fitted to the log glucose concentration
this research,
the investigators
and Care as promulgated
sources, National
Academy
was
described by Markowitl rate of
the plasma glucose concentration
the glucose infuslon
R,
the actual
nephrectomy
adhered
by the Committee
of Sciences- National
always
to the
Guide /or
for Laboratory
Research Council.
passed within
Animal
+2”,,
the k of the
Laboraror~~ Anirnul Facilities
and Re-
FALLIBILITY
219
OF IVGTT
actual measured value. The line was extrapolated to give the theoretic glucose concentration when t = 0 (CO): the glucose space (V) was calculated by dividing the quantity of gluyse administered in the bolus by (Ca - C,,), where C,, is the “experimental fasting level,” or the concentration of glucose when the glucose tolerance curve levels off (for an example, see Fig. I). Total body glucose (G,, in mg/kg) was obtained by multiplying V x C,,. The roportionality constant k was calculated from the following equation: C,, - Ccs = (Cc - C,,)e, where C, is the arterial glucose concentration at r = x. The rate of uptake of glucose (Rd) and the R,, which are equal when the animals are in a dynamic steady state (as they were in our experiment), was calculated by multiplying k x G,. The second aspect of our investigation was to compare results obtained in intact animals by means of the Hlad et al. technique’ with those obtained by means of the primed-constant infusion of 6-3H-glucose (New England Nuclear, Boston, Mass.). In order to assess the applicability of the Hlad et al. technique to the study of glucose kinetics during stress or trauma. we performed studies in both normal guinea pigs and guinea pigs infused with live Escherichia co/i. The guinea pigs had chronic catheters (PE 50) surgically implanted (halothane anesthesia) in the carotid artery and jugular vein 48 hr before the experiment. The animals were fasted from the time of surgery until the start of the experiment. but were provided with water ad libitum. All experiments were performed in unanesthetized. unrestrained animals. In one group of animals (n = 4). we compared the values obtained for R, by means of the IVGTT and by means of the primed-constant infusion of 6-3H-glucose both before and 4 hr after the infusion of approximately IO lo live E. coli bacteria. The procedure for determining R, by means of the primed-constant infusion of radiolabeled glucose has been in use for many years,’ and we have validated this technique experimentally in the hepatectomized, nephrectomized dog preparation used in this study.’ The selection of 6-3H-glucose as an appropriate tracer molecule has been dealt with by Altzuler et a1.s Between 14 and 2 hr after the start of the 6-3H-glucose infusion. three blood samples (0.2 ml) were drawn. and the plasma glucose specific radioactivity was determined.g Since the animals were in an isotopic steady state by that time, R, was calculated by dividing the isotope infusion rate by the mean plasma glucose specihc activity.9 Then 500 mg/kg of unlabeled glucose was administered intravenously as a bolus; arterial blood samples (0.07 ml) were drawn at 5. 7.5, IO. 15, 20, and 30 min after the unlabeled glucose infusion and analyzed for glucose concentration. Three hours were then allowed for the animal’s blood glucose to stabilize at the original level, after which the bolus of live E. coii bacteria was injected through the arterial catheter without interruption of the 6-3H-glucose infusion. At 3% and 4 hr after the E. co/i injection. two blood samples were drawn to determine R, by means of the radiolabeled tracer technique.’ Another bolus of unlabeled glucose (500 mg/kg) was then injected intravenously, with samples again being drawn at 5, IO, 15. 20. and 30 min in order to calculate R,. k, and V according to Hlad et al.’ The time course of the elfects of E. co/i injection on R, determined by means of the primedconstant infusion of 6-‘H-glucose was studied in six guinea pigs. After isotopic steady state had been attained and basal R, determined. these animals were infused with approximately IO’” live E. co/i. Arterial samples were then drawn every hour and R, and Rd were calculated as described before.9 The suppressibility of endogenous glucose production by an exogenous infusion of unlabeled glucose was evaluated in a final group (n = 5). The same basic protocol was used as described above. except that a constant infusion of unlabeled glucose was started and maintained at the rate of IO mg/kg/min between the second and fourth hour postinfusion of E. coli. An isotopic steady state was achieved by I$ hr after the start of the unlabeled infusion, and endogenous R, was determined during the next 30 min of the infusion by means of the radiolabeled tracer technique.g We had previously found this rate and duration of exogenous glucose infusion to lower IO endogenous R, 67”,, in normal guinea pigs. RESULTS
The data from the glucose tolerance three hepatectomized, nephrectomized
test (IVGTT) administered to one of the dogs are presented in Fig. 1. When the
WOLFE,
ALLSOP,
AND
BURKE
oo( I-
5a )3ccP
ICKI-
5c )L
3c I-
!B, I80
205
230
255
280
Fig. 1. (A) Plasma glucose concentration in naphrectomized, hepatectomired dog (No. 3) receiving Q constant infusion of glucose (2.05 mg/kg/ min) from t = 0 to t = 275 min. The IVGTT (300 mg/kg) was started at t = 155 min. (B) IVGTT data from dog 3 plotted on semilogarithmic fasting level” of glucose. scale. C,. . “experimental
data were plotted on a semilogarithmic scale (Fig. I), Co was found to be 330 mg/dl. C,, was 57 mg/dl, although the plasma glucose concentration was 87 mg/dl before the administration of the IVGTT. For this dog, k was calculated to be O.O278/min, V was 110 ml/kg, and R, was 3.06 mg/kg/min. The calculated R, was therefore 49”,, higher than the actual R, of 2.05 mg/kg/min. The values obtained from the other dogs were similar in most respects (Table I), although the calculated value for R, was overestimated to an even greater extent in the other dogs. In order to calculate the basal glucose kinetics from an IVGTT administered to guinea pigs. we curve-fitted a line to the mean blood glucose concentration for the group obtained at each time interval after the bolus of glucose. This modification of the technique of Hlad et al.’ has been used frequently when the IVGTT have been administered to small animals.” The resulting curves obtained before and 4 hr after the infusion of live E. co/i are presented in Figs. 2 and 3. The values calculated on the basis of these curves are presented in Table 2. The most striking finding was the impairment of “glucose tolerance” after the E. coli infusion, reflected by a decrease in k from O.O54/min before the E. Table 1, Glucose Kinetics Calculated KIW,Vn Animal
From IVGTT in Hepatectomired,
Proportionality Constant
R,*
(min-‘)
(k)
Glucose Size
NO.
(w/kg/minj
1
2.05
.0317
117
2
2.05
.0290
3
2.05
.0278
*R,: rate of appearance
of glucose.
W/kg)
Nephrectomized
Pool
Dogs
Error Calculated f%
of
Calcvloted R,
(%I
3.18
+55
125
3.66
+79
110
3.06
t49
221
FALLlBlllTY OF IVGTT
r
1oot
Mean plasma gluFig. 2. mse concentration following injection of 500 mg/kg gluinto normal, fasting case guinea pigs (N = 4).
I Oi
coli infusion to O.O068/min at 4 hr after the infusion. These changes in k value corresponded to a reduction in R, from 17.36 mg/kg/min before the E. coli infusion to 3.59 mg/kg/min after E. cob. In contrast, R, determined by the primed-constant infusion of 6-3H-glucose (Ra3H) revealed a mean value of 4.81 mg/kg/min before E. coli and increased to 9.64 mg/kg/min 4 hr after the E. cofi infusion (Table 3). The results of the time course study of R,-‘H in six guinea pigs infused with E. coli are presented in Fig. 4. During the first 6 hr after the E. co/i infusion, both Ra3H and R:H were always significantly elevated (according to single
Fig. 3. Mean plasma glucose concentration following injection of 500 mg/kg glucose into fasting guinea pigs (N = 4) that had been infused with 1O’O live C. coli 4 hr earlier.
I
I/I
0
5
10
15
20
25
30
MII’JLJ TES
Table 2. Glucose Kinetics Calculated
From IVGTT in Guinea
Proportionality
GllJCOSe
Constant (k)
Pool Size (V)
(mine’)
(44
Pigs GlWXe
Turnover (R,) (m/Wmin)
Control
0.054
255
17.36
4 hr post E. co/i
0.0068
285
3.59
Values
represent
calculations
from
pooled
data
from
four
animals.
WOLFE,
222
Table
3.
Individual
Rates
Means
of Appearance
of the
of Glucose
Primed-Constant
(R,)
Infusion
in Guinea
ALLSOP,
Pigs
AND
Determined
BURKE
by
of 6-3H-Glucose R, (w/kg/min)
Animal
4 hr COiTtd
No.
Studies
were
Post E. Co11
1
5.42
2
3.69
7.03
3
4.43
10.61
4
5.68
11.09
Mean
4.81
9.44
done
in the some animals
in which
IVGTT
9.81
were
administered
(Table
2).
factor analysis of variance and Dunnet’s t test, p < 0.05). During the first 4 hr, the plasma glucose concentration rose because Ra3H exceeded Rd3H. Between 4 and 6 hr after E. coli infusion (the time corresponding to when the IVGTT were given), Re3H remained significantly elevated but the plasma glucose concentration fell slightly, indicating an even greater increase in Rd3H. When glucose was infused at the rate of 10 mg/kg/min for 2 hr into animals given live E. cofi, there was no suppression of endogenous R,-‘H whatsoever and the plasma glucose concentration rose to a mean value in excess of 500 mg/dl (Fig. 5). We have previously shown that the same procedure suppressed endogenous Ra3H 67”,, in normal guinea pigs.”
r
i
Fig.
4.
Rate
(R,),
ance
(R.j),
cose
concentration
the
rate
of
ance
and
infusion
of
lOlo
live
were
determined
the
of
E.
appear-
disappear-
plasma
approximately
co/i.
R, by
and means
primed-constant
of 6-3H-glucose. of individual
glu-
following
data
Rd of
infusion
Mean
f
(N
= 6).
SEM
FALLIBILITY
223
OF IVGTT
Fig. 5. Plasma glucose conendogenout centration and rate of appearonce of glucose (R,) in experimental sepsis before and during (I Z-hr infusion of glucose (10 mg/kg/ min) as compared to control value. R, was determined by means of the primed-constant 6-3H-glucose. infusion of Mean f SEM of individual data (N = 5).
L
100
0
POST E.coli
POST E,coli GLUCOSE
DISCUSSION
A number of assumptions must be made in order to calculate the rate of production of glucose (R,) from a glucose tolerance test. Fundamental among these is the belief that it is valid to curve-fit data from a glucose tolerance test and that each term in the mathematic description of that curve has a physical correlate such as the glucose pool size or the rate of tissue uptake of glucose. Other assumptions made by Hlad et al.’ are (1) that there is a single glucose space in which a bolus of injected glucose is uniformly mixed within 10 mitt, (2) that the tissue uptake of glucose is linear from 100 to 600 mg/dl. (3) that all disappearance of glucose from the sampled glucose pool is attributable to the tissue uptake of glucose, and (4) that an intravenous bolus infusion of glucose does not disturb the basal glucose kinetics. Other mathematic analyses of the IVGTT share many of these assumptions, although some writers assume that the glucose bolus reduces endogenous glucose production to zero,4 so that the k value is taken to represent tissue uptake only rather than the net balance between tissue uptake and continued production. This alternative method of analysis4 is addressed in the discussion of the guinea pig data. The prediction of R, from IVGTT in hepatectomized, nephrectomized dogs receiving a constant infusion of unlabeled glucose would be expected to be more reliable than doing so in normal dogs because, contrary to physiologic responses, ‘* our dog preparation complied with two of the assumptions of Hlad et al:’ first, the basal R, was not affected by the bolus injection of glucose, and second, in the nephrectomized dog there are no nonlinear losses of glucose in urine due to tubular reabsorption capacity being exceeded during an IVGTT.13 The inability to predict R, accurately in a system “loaded” to comply with assumptions is therefore striking evidence that the IVGTT and the equation of Hlad and associates cannot reliably measure R,. It should be emphasized that
224
WOLFE,
ALLSOP,
AND BURKE
in six studies using the same animal model, the primed-constant infusion of 6-3H-glucose measured R, with a coefficient of variation of 3.5”,, about the true value.5 The question arises as to why the Hlad equation fails to predict R, in a model designed to comply with initial assumptions. The work of Steele and associates14 as well as our own’ has established that the volume of distribution of glucose consists of at least two theoretic pools. Steelei conceptualized these pools as consisting of a fast-mixing compartment, which our isotope studies indicate corresponds roughly to the plasma volume.5 and a slow-mixing pool equal to the remainder of the glucose space, which is presumably interstitial fluid. Thus, the rapid decay of glucose concentration during the first IO min after the bolus injection of glucose probably represents mixing in the plasma compartment. The rate of decay in plasma glucose concentration (li/min) after the completion of mixing in the plasma compartment is then determined by the balance between the rate of diffusion of glucose from the plasma to interstitial fluid and the rate of continuing glucose production. The rate of diffusion from plasma to interstitial fluid following an IVGTT will depend on the concentration gradient across the capillary wall and any further concentration gradients within the interstitial fluid compartment. The rate of tissue uptake of glucose is one factor influencing the interstitial fluid glucose concentration but is not the sole factor determining the gradient between plasma and interstitial fluid. For example, the percentage increment in plasma glucose concentration and hence diffusion gradient will depend on the size of the glucose bolus relative to the plasma glucose concentration at the time it is injected. On this basis a standard glucose bolus in a hyperglycemic situation would establish a smaller percent increment in plasma glucose. a smaller diffusion gradient between plasma and interstitial fluid, and hence a smaller k value. This argument agrees well with the relationship between the k value and total glucose pool size observed by Wilmore et al.3 It also explains the observation of CahillI that when a glucose tolerance test is repeated in a diabetic who has no insulin response, the k value is inversely proportional to the blood glucose concentration at the start of each test. The failure to correctly calculate R, by the Hlad technique in our dogs supports the argument that factors other than rate of tissue uptake significantly influence the k value even when R, is kept constant. In intact animals the application of IVGTT to measure R, is further compounded by the unknown extent of suppression of endogenous glucose production by the glucose bolus. The evidence in this paper and previously published results” indicate that exogenous glucose infused at IO mg/kg/min depresses endogenous production approximately 67”,, in normal guinea pigs and not at all in guinea pigs given live E. cofi. The suppressibility of gluconeogenesis after of R, from IVGTT in burn injury is again different. lo The 250”,, overestimate normal guinea pigs can be attributed in part to the assumption of Hlad et al. that endogenous R, proceeds uninterrupted after the glucose bolus. However. if we had not made that assumption and therefore did not use the C,, value in the calculations, the R, would still have been overestimated by 67”,,. On the other hand, the failure to use C,, in the calculations for the septic animals
FALLlBlllTY
OF IVGTT
225
magnified the already large error encountered when the C,, value was used. The lack of suppressibility of endogenous R, in the septic animals probably accounted for the desirability of using the C,, value in that situation. Unfortunately, there is no way to know a priori whether the use of the C,, value will improve or decrease the accuracy of the calculated R,, and in any event substantial errors persist regardless of its inclusion or deletion. The errors in calculated R, may also be partly due to the unfounded assumption that mathematic terms, such as the k value, that describe successive data points have physical correlates. Berlin and associates” have pointed out that a single exponential curve can approximate a process that is in fact a two-compartment constant coefficient system containing no less than four different turnover rates. The danger of assigning physical characteristics to the terms used in the mathematic descriptions of the data is increased by the fact that with IVGTT the data points frequently lie further from the best fit monoexponential curve that can be explained by experimental error. Thus, in intact animals, the combination of an undefined diffusion gradient, a variable suppressibility of endogenous R,, and difficulties in equating mathematic terms with physiologic events can result in an R, predicted from an IVGTT that has no relationship to the actual R,. For example, using the validated isotopic technique’ R, was elevated loo”,, after E. coli. yet when calculated by the IVGTT method R, was depressed 80”,, after E. coli. A final comment is required about the use of pool size and k value as isolated indices of glucose kinetics in vivo. Fortuitously, in the calculation of R, the error in k is offset to some extent by the error in glucose pool size, which is determined by extrapolation of k. It would therefore appear that the glucose pool size and k value considered separately are even less reliable indices of glucose metabolism than is the R, calculated from them. This is exemplified by the results presented in Fig. 1. With only one exception, the data points from 10 to 70 min after the glucose injection fell on a single exponential curve, yet when that curve was extrapolated to t = 0 in order to calculate the total glucose space (V), a value was obtained that was less than one-half the value originally established by Wick et al.” as the glucose space of hepatectomized, nephrectomized dogs. The unreliability of the k value is particularly noteworthy, since the k value is widely believed to be synonymous with the rate of tissue uptake of glucose. The flat curves we observed in septic guinea pigs (Fig. 3) are almost identical to those that have been reported following several forms of trauma.““’ Such “diabetic” IVGTT curves have classically been used to diagnose “insulin resistance”-an inability of tissues to take up glucose after trauma.” However, the validated radioisotope data reveal that the tissues can actually be taking up glucose at twice the normal rate at a time when the k value is decreased markedly. The conflict of these findings with the traditional interpretation of k results from the fact that the rate of tissue uptake of glucose is only one factor contributing to the k value. The extent of suppression of gluconeogenesis as well as the undefined diffusion gradients within the glucose pools can dominate the shape of the IVGTT, and consequently the k value should not be used as an index of glucose uptake.
226
FALLIBILITY
OF IVGTT
REFERENCES I.
Hlad CJ Jr, Elrich
on the
kinetics
Invest 35: I I39 2. Wilmore
I 149,
1956
DW.
Mason
in
glucose
Jr: Alterations thermal
H, Witten
of glucose
TA: Studies
utilization.
3. Wilmore
DW.
Mason
glucose
flow
with gram-negative 143:720
Jr. Pruitt
kinetics
BA
following
AD
Jr. Pruitt
in burned
the intravenous 23:103-l
BA
Obstet
D: Glucose
assimila-
evaluation
of
glucose tolerance
test. Clin
Sci
reliability
Wolfe
RR,
Burke
Jl-:
of rates of glucose appearance
calculated
from constant
tracer
The
in vivo
infusions.
Bio-
them J (in press) 6. Markowitz
J. Yates
WM.
A simple one stage technique in thedog.
Burrows
J Lab Clin Med IX:1271
fect of insulin on utilization glucose.
Am
and production
by a new tracer
01
1X9:43 50.
14. Steele R.
Glucose
Barkai
turnover
with various
tn the
burn
guinea
pigs.
dog
C. et al: obtained
glucose.
Am
Glucose Phyrml
Miller
injury Surg
HI. on
dtabetic
method.
dogs
Metabolism
of the Kidney
studies
in
and
Year Book. 1971
hepatic
R: Does
sugar output’!
evrscerated
dogs.
a
14C
Am
J
on pools. Fed Proc
679. I Y6-t Physiology
Diabetes 20:7X5 17. Berlin
N.
Berman
The application Med 6X:423
of insulin tn man.
799. I97 I M.
Berk
PO,
ofmulticompartmental
of cltnical
medictne.
et al:
analv\ts Ann
Intern
436. 196X AH.
Drury
DR.
Makay
EM:
Glu-
cose space of the body. .Am J Phystol 163:224 27x. 1950
Elaht D. et al: Efglucose
Gynecol
Carbo-
infection.
197:60 62. 1959
IX. Wick J
1667. 1975
RR.
fect of 364, I977
values
Bjerknes
species of labeled
Phvsiol229:1662 9. Wolfe
A.
233:E80-
Jr:
Btshop JS. Levine
glucose load inhibit
to problems N.
burn
and re-
Jr: Glucoregula-
and
Physiology
16. Cahill GF.
1957 8. Altzuler
of
372, 1971
Body Fluids. Chicago,
I
AD
G
in normal
13. Pitts RF-
22:67
127X. 1933
J Physiol
Mason
15. Steele R: Retlections
WH.
for hepatectomy
7. Wall JS. Steele R. de Bodo RC. et al: Efcirculating
tlfect
oxidation
in Pseudomonas
JS. Hetenyt
responses
20:360
A critical
JR,
Jl-:
Biomed Rea 2:C I 67. 196X
Comput
tory
Kl-.
metabolism
12. Cowan
13. 1962
5. Allsop
LaNoue
recorded
V. Marrack
Burke
E85, 1977 II.
patients
sepsis. Surg Gynecol
tion in hyperinsulinism.
RR,
on glucose turnover.
hydrate
83. 1975
724, 1976
4. Marks
IO. Wolfe trauma
cycling in guinea pigs. Am J Physiol AD
injury. Surg Forum 25:81
Jr: Impaired
J Clin
turnover
Obstet
In
144:359--
IY.
Frayn
KN.
sulin secretion the rat in viva. 1975
Eti’ects ol burn injury on in-
and on sensttivity Eur J Clin
to insultn
Invest
5:33l
in
337.
